MIDI-mapper for OpenSpace (or any javascript API)

With the help of Claude.ai, I've put up a MIDI-mapper for OpenSpace at OpenSpace MIDI mapping.

The readme.txt lists the prompts used, and usage. Copy-pasting from my post on OpenSpace Slack: With this, we can control OpenSpace using any MIDI control surface or MIDI instrument. Interesting effects like speeding up time for one (MIDI) violin string, slowing down for another string, etc can be thought of.

I've used our KORG nanokontrol - earlier version of this product - to do the development and testing.

By adding suitable functions, this could be extended to call any javascript API using a MIDI controller or MIDI instrument.

Bosch Wiper blades for Hyundai Santro

 PB helped me out with the following info:

Attached is the wiper blades size for Santro
List shows 20" on Driver side and 16" on Passenger side

TYPES OF BOSCH WIPERS
Bosch Clear Advantage — a flat-blade / beam-style wiper. Designed to give uniform pressure across the windshield, better contact, smoother wiping, and more streak-free clearing (especially good in heavy rain or at high speeds).

Bosch ECO (conventional / bracket-frame wiper) — a more traditional wiper with a metal (or mixed) frame + rubber blade. Usually more budget-friendly and covers basic everyday wiping needs.

AMAZON LINKS
Bosch Clear Advantage 16"

Bosch Clear Advantage 20"

Bosch ECO 16"

Bosch ECO 20"

I have been using Bosch ECO, and it works well. 

Will try Clear Advantage next time.

And, found that there were video guides on how to fit the blades, in the amazon listings. Also available on youtube, at Bosch ECO and Clear Advantage.

On my (new) Santro, the 20" blade seems a bit short. Perhaps I'll try the 21-inch blade next time.

hardware accelerated video encoding on Intel Iris graphics has issues with jpg frames input - resolved

 From ChatGPT,

ffmpeg -i filename_%05d.jpg -c:v hevc_qsv -preset veryslow -global_quality 0 output_lossless_qsv.mkv

For 4K (4096x4096) files, this was running at around 8 fps on my i5 laptop, as against 2 fps for non-accelerated libx264. The mistake I made earlier was to put the preset before the -c:v.

• -i must come before input options end.
• All encoder options (-c:v, -preset, -global_quality) must be placed after the input and before the output filename.

If you place -preset before -i filename_%05d.jpg, FFmpeg thinks it’s a decoder option, causing the error:

Codec AVOption preset ... is not a decoding option.

With audio, we would need something like

ffmpeg -r 30 -i KG-LS_%05d.jpg -i audio.wav \

  -c:v hevc_qsv -preset veryslow -global_quality 0 \

  -c:a aac -b:a 192k \

  -shortest \

  output_with_audio.mkv

# or on Windows,

ffmpeg -r 30 -i KG-LS_%05d.jpg -i audio.wav ^

  -c:v hevc_qsv -preset veryslow -global_quality 0 ^

  -c:a aac -b:a 192k ^

  -shortest ^

  output_with_audio.mkv

Edit - this produces a low-bitrate video with colour issues on VLC (and perhaps on other players. The python front-end presets, like the command below, might be useful - 

FFmpeg command to copy-paste:

ffmpeg -y -i in_%05d.jpg -i map_x_directp2.pgm -i map_y_directp2.pgm -i weight_alpha_mask.png -filter_complex '[0:v][1:v][2:v]remap[remapped];[3:v]format=gray,scale=4096:4096,colorchannelmixer=rr=1:gg=1:bb=1[mask_rgb];[remapped][mask_rgb]blend=all_mode=multiply[blended];[blended]scale=3840:2160,setsar=1,setdar=16/9,format=yuv420p[out]' -map '[out]' -map '0:a?' -c:v hevc_qsv -preset fast -crf 23 -pix_fmt yuv420p -c:a aac -b:a 128k D:/out-W4K.mp4

So, without warping, that would be - 

ffmpeg -y -i in_%05d.jpg -i audio.wav -c:v hevc_qsv -preset veryslow -crf 1 -pix_fmt yuv420p -c:a aac -b:a 128k out.mp4

But unfortunately, even the warping command produces a low bitrate video with colour issues. Green tinted instead of purple tinted for the earlier videos. Even changing the codec to libx265 did not help. So, some incompatibility in the input frames is suspected. Maybe I'll just use avidemux instead instead of trying to troubleshoot.

From ffmpeg -h

Getting help:

    -h      -- print basic options

    -h long -- print more options

    -h full -- print all options (including all format and codec specific options, very long)

    -h type=name -- print all options for the named decoder/encoder/demuxer/muxer/filter/bsf/protocol

ffmpeg -h encoder=hevc_qsv
....

  -preset            <int>        E..V....... (from 1 to 7) (default medium)

     veryfast        7            E..V.......

     faster          6            E..V.......

     fast            5            E..V.......

     medium          4            E..V.......

     slow            3            E..V.......

     slower          2            E..V.......

     veryslow        1            E..V.......

....

  -profile           <int>        E..V....... (from 0 to INT_MAX) (default unknown)

     unknown         0            E..V.......

     main            1            E..V.......

     main10          2            E..V.......

     mainsp          3            E..V.......

     rext            4            E..V.......

     scc             9            E..V.......

    

Claude.ai pointed out that the colour tint issues are probably due to wrong pixel format, and the low bitrate is probably because "QSV's CRF implementation is unreliable". So, suggestion is to use nv12 pixel format and global_quality instead of CRF, where quality of 15-20 is near lossless. Or force a high bitrate if that doesn't work - -b:v 20M -maxrate 25M

ffmpeg -y -r 30 -i in_%05d.jpg -i audio.wav -c:v hevc_qsv -preset veryslow -global_quality 18 -pix_fmt nv12 -c:a aac -b:a 128k out.mp4

And that solved the problem.

No colour tint, hevc_qsv encoder resulted in a 25 Mbps video, libx265 encoder with the same preset and quality settings but without the -pix_fmt nv12 resulted in a 7 Mbps video, both of good quality.

Brevo transactional emails and DNS setup.

After email discussions spanning a few days, I set up google workspace accounts and added a couple of subdomains to google workspace - Brevo's setup includes CNAMEs for DMARC and DKIM records in the format subdomain1.ourdomain.org is a cname for b2.vsubdomain1-ourdomain-org.dkim.brevo.com and so on. 

So, set up email1@subdomain1.ourdomain.org as an alias for email1@ourdomain.org and so on, pointing the MX records to google workspace, so that bounce messages would be able to go somewhere. The reason I'm just not creating email ids like email1@subdomain1.ourdomain.org without aliases is because Brevo uses DMARC (a method of email server authentication) which would cause emails sent from google's servers to be marked as spam if we set Brevo to send authenticated emails from those email ids.

There were lots of back-and-forth emails saying emails are not being delivered. Finally, the dev team set up the domain authentication once again (which they had not done for subdomain2.ourdomain.org), sent me the details of the cnames & txt records to be added, I added them, and then all was OK.  

strategies for dealing with Google Workspace account filling up

Copy-pasting from an email:

Here are some ways to clear up space in <google workspace account>

1. From https://support.google.com/a/users/answer/14300711

You can go to drive.google.com/drive/quota. There, you see how much storage you use in Google Drive, Gmail, and Google Photos. It also lists the largest files in your google drive, which you can delete if not needed.

2. If GMail is the service which is using up most of the storage, then you can consider deleting files based on how large the attachments are, or based on how old the email is.

To find the emails with attachments greater than 10 MB, enter the following in the search box at the top of your GMail - 

has:attachment larger:10Mb

and hit Enter. 

Similarly for attachments greater than 5 MB etc. 

(You can also use the advanced search link on top of your GMail to achieve this.)

For finding emails older than a particular date, there is a drop-down, "Any time" just below the search email box. To find all emails earlier than 2015, say, you can first clear the "Search mail" box and then choose the "Any time" drop-down, and choose custom, like the screenshot below.

Using these search methods, you can try to delete unwanted emails, spam, internet forwards, outdated correspondence, and so on.

Once emails are deleted, you will then need to go to "Trash" and choose "Empty Trash Now" if you want the space savings to be applied ASAP. Even after this, Google might take some time before the % used decreases.

In case you are not able to free up space in this way, and need to back up old correspondence, here are some ways of doing that, by forwarding to another account, or by using a product called mailstore, etc, details in the link below - 

https://hnsws.blogspot.com/2019/04/mailstore-for-achiving-emails.html

Send user digests - Moodle task fails - sends hundreds of emails to Admin

On one of our Moodle instances, emails are supposed to go out through Google Workspace using Google SMTP servers. That being the case, we can't send an unlimited number of emails. So, email notifications had been globally - Site Administration > General > Messaging > Notification settings, which is located at server.url/admin/message.php

When there was a request to allow emails of a particular type to particular users, after enabling the email notification plugin at the above page, the next time a forum post was made, there was a cascade of Send user digests failed and Send user notifications failed messages.

For the immediate failure messages to stop, deleted all the adhoc tasks from mod_forum,

delete from prefix_task_adhoc  where prefix_task_adhoc.component = 'mod_forum';

I had previously set emailstop=1 on the database users table, for all users except the handful which should receive emails. Possibly the reason for the failures was that forcesubscribe=1 was set for many forums for many / all users. 

So, now I set it to zero (email disabled) for all forums on this instance, with

update prefix_forum set prefix_forum.forcesubscribe = 0;

and also updated the defaults,

Site administration > plugins > plugins overview > (chose mod_forum) > settings >

site.url/admin/settings.php?section=modsettingforum

and set subscription mode = subscription disabled for both defaults,

forum_announcementsubscription

and

forum_subscription 

Also, went to prefix_forum_subscriptions table and deleted all entries - which I suppose would be various users' subscriptions to various forums.

Hopefully this is sufficient to prevent the hundreds of failed send notification tasks. 

Update 13 Nov - Again received hundreds of task failed messages. Asked Claude.ai, and made the following changes as suggested, in addition to the above.

# Individual discussion subscriptions might still exist
DELETE FROM prefix_forum_digests;
DELETE FROM prefix_forum_discussion_subs;

UPDATE prefix_user SET maildigest = 0 WHERE emailstop = 1;

# Users might be getting auto-subscribed when they post
UPDATE prefix_forum SET trackingtype = 0;

TRUNCATE prefix_forum_track_prefs;

run 'Issue Certificates' task more often

There was a request for setting up email notifications for one Moodle course, a particular user to be notified when any user completes that course. I had earlier set many tasks to run less often due to our server being overloaded. 'Issue Certificates' task was one of them, especially since it needed some optimizations as per this thread. Checking the logs, I could see that issue certificates was taking between 30 and 120 seconds - so, increased the scheduled task cron frequency from once a day to once an hour, for two of our Moodle instances. 

Set to run every hour at the 30th minute (8:30, 9:30 etc) on one instance and at the 45th minute (8:45, 9:45, etc) on the other instance, in Site Administration > Server > Tasks > Scheduled tasks > \mod_customcert\task\issue_certificates_task 

installing python modules on Linux and Windows

For trying out LLM-generated python scripts for "warping", (still works in progress)

https://github.com/hn-88/python-ffmpeg-warp

https://github.com/hn-88/realtime_video_warper

https://github.com/hn-88/realtime_warp_c

On Windows, various python modules can be installed with 

pip install modulename

where we sometimes have to do an internet search to find the exact modulename. 

Eg. How to Fix "ModuleNotFoundError: No Module Named 'PIL'" in Python

pip install pillow

On my Linux Mint machine, where I've installed python using the default deb package with apt, 

ModuleNotFoundError: No module named 'imageio'

and when trying to install with pip,

× This environment is externally managed

╰─> To install Python packages system-wide, try apt install

    python3-xyz, where xyz is the package you are trying to

    install.

sudo apt install python3-imageio # this worked.

proctoring plugins for Moodle

Copy-pasting from an email exchange:

 ... would require us to conduct a proctored test. Moodle has several plug-ins. They are promoting 'Proctoring for Moodle'. I also find Quilgo to be good. Can we install these 2 plug-ins (and any other plug-in recommended by you) in (our dev server) for us to test them out?

Here is a list of proctoring plugins for Moodle, probably you would have seen this already. https://moodle.org/plugins/?q=proctor

'Proctoring for Moodle' is free software, but if you want to use the automated face-recognition, they ask us to contact them. "Moodle Proctoring plugin is a Quiz Access plugin designed to capture a user's picture via webcam to identify the individual attempting the Moodle quiz."

A demo is at https://www.youtube.com/watch?v=NcrE1OIj6_Q

In that video, it is mentioned that features like capturing student's screen for "suspicious behaviour" like copy-pasting into Google etc, requires Proctoring Pro.

But the "purchase" Proctoring Pro link seems to be broken. -

https://proctoring.elearning23.com/ 
https://elearning23.com/moodle-proctoring-pro-details/
https://elearning23.com/wp-content/uploads/2024/10/Proctoring-Pro-v2.0.0-User-Manual.pdf

(page not found etc)

"Quilgo® Proctoring for Moodle is a highly scalable plugin that allows you to include a camera, screen, track activity, and receive reports on test completion. Online assessments become easier and more reliable. Easy to use, no API customization or maintenance required." But the free plan is only for 20 tests, and the cheapest paid plan is $1200 for a year or $140 for a month.

https://quilgo.com/products/moodle 

A demo/tutorial is at https://www.youtube.com/watch?v=-pwwImIprSM

I have installed 'Proctoring for Moodle' and "Quilgo® Proctoring for Moodle" on (our dev server) now.

can QUIC be added to our dot net app

There was a request from the developers asking if we can enable QUIC support to the back end of an app for which we had disabled QUIC at cloudflare to prevent issues.

My reply was,

QUIC on Apache web server (which we use as the reverse proxy for dotnet) is experimental. 

https://www.linuxbash.sh/post/configuring-http3-quic-with-apache

We would have to use the experimental module, or shift to another web server like nginx etc. 

Since migration to a different web server is not simple, and the QUIC module is experimental on Apache, I would recommend against enabling QUIC for now. 

Also, QUIC security headaches for admins - 

https://www.networkstraining.com/what-is-quic-protocol/

handwriting recognition - ChatGPT is still the best

Uploading a page from my diary and checking how good the various LLMs are at handwriting recognition - ChatGPT is good for regular snaps.

ChatGPT -

JUN 7 FRI

 Social service instead of jogging as you're coming – making the Ashoka trees straight.

 Afternoon no class – we went to Mandir ~ 1:45, You came ~ 3:15. Vedaṁ, balloons, exerts, song – W/N pg writing & study this time Krishnamoorthy, IMI – problems in C...

Claude had done well at coding - better than ChatGPT and Gemini, but fared worse in handwriting recognition - Claude's version -

JUN 7 FRI

Bowl Service instead of dragging so forth coming - making the little trees sharply. Afternoon no class - we went to Nandu ~ 1:45, Su came ~ 3:15. Yadav, balloons, esotts, songs N/N pg writing & study his tone. Kishnamathy I/W - problems in C...

My corrected version - 

Social service instead of jogging as You're coming - making the Ashoka trees straight. Afternoon no class - we went to Mandir ~ 1:45 pm, You came ~ 3:15 pm. Vedam, balloons, escorts, song – WIN prg  during waiting at Mandir & study hours time. Krishnamoorthy, 1st MSc - problems in C...

Edit - 9 Nov - But when I edited together 5 pages to form a single long page, ChatGPT started hallucinating instead of transcribing!

Excerpt from ChatGPT's version of July 5-14 - 

JUL 5, FRI

 Toured Siddalingeshwar Plantation.

 Found good crop condition — 4½ acres of Amlan.

 Evening went to A.P.S. meeting at old students’ association.

 9:45 P.M. came home.

 Supper.

Actual corrected version,

"Pseudo heterodyning" - frustrating as could get only π phase change. Used signal analyzer to acquire data. S⁴ again seminars on 902...

Timetable again reverting to old schedule, 9 to 3, from Monday.

claude is good at coding

Pleasantly surprised to find that Claude.ai, running the Sonnet 4.5 model, with just a free plan and a single prompt, created a usable UI for a timeline editor in html / javascript, when supplied with javascript helper functions copy-pasted along with the prompt - the initial commit is at

https://github.com/hn-88/openspace-scripts/commit/79ff8dfcea16d7df944d1675f8ec3c6b4b9f3bd8

Edit - Subsequent commits were needed to make it actually functional, connect to OpenSpace, etc. using code copied from Jeff Nee,

https://github.com/hn-88/openspace-scripts/blob/main/OpenSpace-ConstellationsTonight_draft/00-readme.txt

The "finished product" is here,

https://github.com/hn-88/openspace-scripts/tree/main/OpenSpace-Keyframe-Timeline-editor

slerp javascript

via

https://stackoverflow.com/questions/68945391/how-do-i-get-a-smooth-interpolation-between-keyframes

https://github.com/mrdoob/three.js

MIT licensed Math library - for SLERP interpolation for moving the OpenSpace camera smoothly.

Edit: 2026 - Claude.ai generated all the necessary slerp interpolation by itself, so I did not have to use three.js

python QT UI in VisOCT-Explorer

Via email notification from researchgate about this preprint, https://doi.org/10.1101/2025.10.01.679626

https://github.com/FOIL-NU/VisOCT-Explorer

uses python QT gui,

https://wiki.qt.io/PySide_Binaries_Linux

Of course, for simple UIs, we can just use python's default UI, tkinter, as in https://github.com/hn-88/python-ffmpeg-warp/blob/main/py-ffmpeg-warp.py especially when AI helps us to make the layout with just one prompt.

https://tkinterbuilder.com/comparison.html

Moodle Course Completion

Extracts from an email thread, S asked the question and answered himself - 

...we have a custom report 'Course Completion Status' in one of our Moodle instances.

There is a field 'Course Completion - Completed' and this field should show 'Yes' when the course is completed and 'No' when the user is still doing the course. And then there is another field, 'Course Completion.Student Progress' and this field shows the percentage progress.

For (one particular) course many teachers have completed the course (the percentage is 100); certificates have been issued to them. However, the field 'Course completion - completed' is set to 'No' and not 'Yes'. What could be the problem? Even the system report 'Activity Completion' has all the activities ticked. 

Found out the issue - within the course, the course completion criteria weren't set up. To mark the course as completed, at least one criterion should be enabled. 

In the course page - More->Course Completion, enable one or more conditions. I have enabled 'Custom Certificate-Course Certificate'. And immediately, all the people who have received certificates have been marked as 'Completed'. 

hidden restrictions in Airtel's "unlimited 5G" plans

According to this youtube video, there are some terms and conditions which are shown only in the Airtel app, by which

1. If Airtel thinks it is being used for commercial purposes, they can restrict the "unlimited 5G"
2. If the data transferred exceeds 300 GB in a month, they can consider it "commercial"
3. Tethered use of 5G data - using a mobile hotspot - is not included in the "unlimited 5G"

And one of the comments in that video pointed out that the reason why at some times the 4G data cap gets used up even though 5G is displayed in the status area on the top right of Android phones, is because of Airtel's use of 4G infrastructure to deliver 5G - in which case, the 4G billing gets used. And apparently, the difference can be noticed on Samsung phones with the 5G logo which is displayed. 

It seems 5G with a white background = true 5G.

And 5G with white text = 4G masquerading as 5G.

And this seems to be confirmed by this page, 
https://androidsrc.net/android-samsung-phone-symbols/

Youtube 360 VR videos with a poor internet connection

Most of the time, our office internet connection can't support 2160s resolution for 360 VR videos on Youtube. My current solution is to download the youtube videos using yt-dlp - 

./yt-dlp_linux https://www.youtube.com/watch?v=yHQYd5i7CAE

and then watch.

internet problems with our old wifi router

Found that I had to restart the router often - sometimes more than once a day - when DNS would fail. So, instead of using the router's DNS, decided to use 8.8.8.8 - changed the IPv4 settings on the desktop (Linux Mint - "Advanced Network Configuration" - to use a static ip and added the DNS address. Did not make any changes on the router.

Google Apps Script - get the email of signed in user

There was a request to create a Google Apps Script which would do some spreadsheet lookups based on the email of the user, who would authenticate by signing in to the deployed web app - which would be deployed as having access to "Anyone with a Google account".

Apparently the script would work for anyone having an email id in the same workspace domain, but it would not work for users with generic @gmail.com email ids. 

Session.getActiveuser.getEmail() has the problem that it would need the script to run as the user who is logging in (and not as the creator of the script). Apparently this restriction does not apply for users within the same google workspace - 

https://developers.google.com/apps-script/reference/base/user

"the user's email address is not available in any context that allows a script to run without that user's authorization, like a simple onOpen(e) or onEdit(e) trigger, a custom function in Google Sheets, or a web app deployed to "execute as me" (that is, authorized by the developer instead of the user). However, these restrictions generally do not apply if the developer runs the script themselves or belongs to the same Google Workspace domain as the user."

So, made a workaround - first, a script which will run as the user who is logging in, which will get the email - and then POST it to the second script, which will run as the creator of the script, which will process the information. At the POST step, we could add our own authentication using some salted hash etc. 

The first script would be something like - 

function doGet(e) {

  let userEmail = Session.getActiveUser().getEmail();

  let targetUrl = "https://script.google.com/macros/s/AKfycbyx4CbCNIpgmLd2iaWF4buFh_WZtIyOY3vpwdht_Gillycyflpgu1C2bN-j77P2CD3j/exec"; // Replace with your other script's URL

  let html = `

    <!DOCTYPE html>

    <html>

      <body>

        <p>You are signed in as <b>${userEmail}</b></p>

        <form method="post" action="${targetUrl}" target="_blank">

          <input type="hidden" name="userEmail" value="${userEmail}">

  <!-- add other hidden inputs with salted hash or whatever for extra authentication -->

          <button type="submit">Proceed</button>

        </form>

      </body>

    </html>

  `;

  return HtmlService.createHtmlOutput(html);

}

// With deployment access to "Anyone with a Google account"

// Execute As "The User accessing the web app" 

And the second script would have a doPost() like -

function doPost(e) {

// e.parameter contains form fields

let userEmail = e.parameter.userEmail;

// do whatever other parameter checks for extra authentication

// Call your function to fetch data

let result = getData(userEmail);

let html;

if (result && result.found) {

let data = result.data;

html = `

<!DOCTYPE html>

<html>

<body>

<h3>Data Details</h3>

<p><b>Data1:</b> ${data.Namefield1}</p>

<p><b>Data Num:</b> ${data.Num}</p>

<p><b>Data Program:</b> ${data.Program}</p>

</body>

</html>

`;

} else {

html = `

<!DOCTYPE html>

<html>

<body>

<p>No data found for <b>${userEmail}</b></p>

</body>

</html>

`;

}

return HtmlService.createHtmlOutput(html);

}

Edit - In this implementation, the second script would open in a new tab or new window, with target=_blank when submitting the form - I was asked if it could open in the same window. Tried it, the answer was that no, opening in the same window will not work. Probably because of “X-Frame-Options” or other HTTP headers that block embedding / cross origin operations.

warping with ffmpeg - and some filter_complex examples

Since OCVWarp uses OpenCV remap() and there does exist a remap filter in ffmpeg also, revisited ffmpeg remap in order to make use of the accelerated NVidia codecs, which ffmpeg supports. Asked ChatGPT to generate python code to create the required map files, debugged the code manually, and put the code at

https://github.com/hn-88/OCVWarp/tree/master/python

Currently it's just for the spherical mirror warping of fulldome content, but could potentially be extended to all of OCVWarp's functionality.

This ffmpeg command using the accelerated hevc_nvenc codec runs at 13 fps on a desktop running NVidia RTX 1060 as against 2.5 fps using OCVWarp saving to avc1 codec out.

ffmpeg -i input.mp4 -i map_x_directp2.pgm -i map_y_directp2.pgm -i weight_alpha_mask.png -filter_complex "
  [0:v]scale=3840:2160[scaled];
  [scaled][1:v][2:v]remap[remapped];
  [3:v]format=gray,scale=3840:2160,colorchannelmixer=rr=1:gg=1:bb=1[mask_rgb];
  [remapped][mask_rgb]blend=all_mode=multiply[out]
" -map "[out]" -map 0:a -c:v hevc_nvenc -preset p5 \

-cq 23 -rc vbr -maxrate 15M -bufsize 26M \
-c:a aac -b:a 128k output.mp4

Can potentially even create a UI using Tkinter in python, which would be cross-platform.

Some notes about the ffmpeg commands copy-pasted from the readme there - 

-ss for the start time should come before the input, and 
-t for the duration should come just before the filter_complex - 

For starting at two minutes and stopping after 10 seconds duration,

ffmpeg -ss 00:02:00 -i input.mp4 \

       -i map_x_directp2.pgm \

       -i map_y_directp2.pgm \

        -i weight_alpha_mask.png \

       -t 10 \

-filter_complex "

 [0:v]scale=3840:2160[scaled];

  [scaled][1:v][2:v]remap[remapped];

  [3:v]format=gray,scale=3840:2160,colorchannelmixer=rr=1:gg=1:bb=1[mask_rgb];

  [remapped][mask_rgb]blend=all_mode=multiply[out]

" -map "[out]" -map 0:a -c:v hevc_nvenc -c:a copy output.mp4

And a problematic video was fixed using an initial crop to 4096x4096 (although avidemux could show it as 4096x4096)

ffmpeg  -i "short_nometadata.mp4"        -i map_x_directp2.pgm        -i map_y_directp2.pgm         -i weight_alpha_mask.png      -filter_complex "[0:v]crop=4096:4096[cropped]; [cropped]scale=3840:2160[scaled]; [scaled][1:v][2:v]remap[remapped];[3:v]format=gray,scale=3840:2160,colorchannelmixer=rr=1:gg=1:bb=1[mask_rgb];

  [remapped][mask_rgb]blend=all_mode=multiply[out]

" -map "[out]" -map 0:a -c:v hevc_nvenc -preset p5 -cq 23 -rc vbr -maxrate 15M -bufsize 26M -c:a aac -b:a 128k short_w2.mp4

But unfortunately, the quality of the warped output is slightly lower than what is seen with OCVWarp - slightly less sharp. Perhaps changing the workflow to first remap and then resize (instead of resize to 3840x2160 first and then remap) might improve things.

Edit - 6 October 2025 - Yes - after doing the remap filter first, at the full resolution of the input video, and then the resizing to output resolution brings up the quality to as good or better than OCVWarp. ("Better" when OCVWarp's default avc1 video codec settings don't deliver as good quality as the hevc_nvenc settings above, I guess.) A separate repo has been created for this, incorporating a Python Tkinter UI,

https://github.com/hn-88/python-ffmpeg-warp/

screen capture or screen recording with Linux on NVidia cards

One more way of doing it, in addition to this earlier post which uses a compiled tool gpu-screen-recorder, there also seems to be a script which does ffmpeg capture of a particular window - 

https://github.com/karlomijaljevic/scrd

This seems to use ffmpeg with -x11grab to do the capturing. I've not yet tried it.

dd command to create linux boot disk

via https://linuxvox.com/blog/create-usb-boot-linux-mint/

sudo dd if='/path/to/linuxmint-22-cinnamon-64bit.iso' of=/dev/sdd bs=4M status=progress

2810183680 bytes (2.8 GB, 2.6 GiB) copied, 140 s, 20.1 MB/s

So, on my machine, it took less than three minutes to create a 2.8 GB boot disk.

Fujitsu mini PC not booting

I had a couple of Fujitsu mini PCs purchased cheaply, used from amazon - but neither of them would boot up now, a couple of years after I had used them last. Suspected the BIOS battery - used these videos to help with opening them up - similar, ours was Q920 without the DVD drive - 

Tearing down a Mini PC | Fujitsu Esprimo Q910

Mini PC Teardown - Fujitsu Espromo Q910

Got the CR2032 lithium cell bios batteries also from Amazon, put in new batteries, and they booted up fine. 

Edwiser Bridge Plugin for Moodle

There was a request to install Edwiser Bridge Plugin for Moodle which would enable integration between Moodle and WordPress. 

If this plugin is installed, for it to work, several settings need to be changed - 

https://edwiser.org/documentation/edwiser-bridge-pro/edwiser-bridge-pro/installation-and-user-guide-for-v3-0-0/steps-to-install-the-plugin-for-first-time-users/

I was spooked by some of these requirements, and also due to 

https://vulert.com/vuln-db/CVE-2025-32044

"vulnerability in Moodle that allows unauthenticated users to access sensitive user data through unauthenticated REST API calls. - fixed in 4.5.3"

So, I replied that I would not be responsible for security issues, if any, if this plugin is installed and asked for head of the institution approval if it was to be installed. The request was then withdrawn, so it was not installed.

the find command and mtime - finding disk space used by old files

 Asked ChatGPT, "linux commandline to list total disk space used of all files created or modified inside a directory more than one year ago" and got the following - 

You can do this with find + du or find + stat + awk.
Here’s a one-liner that lists the total disk space used by files in a directory older than one year:

find /path/to/dir -type f -mtime +365 -printf "%s\n" | \
awk '{sum+=$1} END {print sum/1024/1024 " MB"}'

Explanation:

• find /path/to/dir → search inside that directory.

• -type f → only files.

• -mtime +365 → modified more than 365 days ago (roughly one year).

• -printf "%s\n" → print file sizes in bytes.

• awk → sums up sizes and prints total in MB (divide by 1024/1024).

If you prefer human-readable output directly:

find /path/to/dir -type f -mtime +365 -print0 | du --files0-from=- -ch | tail -n1

This gives you a total in du’s human-readable format.

---

 Asked ChatGPT, "tree view of files and directories larger than 1 GB from linux command line" and got the following - 

You can combine find, du, and tree-style output to list only files/directories larger than 1 GB in a tree view. Here are a couple of approaches:

---

1. Using find with tree

find /path/to/search -type f -size +1G -print | tree --fromfile

• find ... -size +1G → finds files > 1 GB

• tree --fromfile → prints a tree structure from the list

If you want directories too (based on contents size), use du instead (see below).

---

2. Using du for directories/files with sizes

du -h --threshold=1G --max-depth=3 /path/to/search | sort -h

• --threshold=1G → only show entries larger than 1 GB

• --max-depth=3 → how deep to go

• sort -h → human-readable sort

This shows sizes and paths (not tree formatting, but size summary is clearer).

---

3. Pretty tree-like display with du

du -h --threshold=1G --max-depth=3 /path/to/search | sort -hr | 
awk '{print $2}' | tree --fromfile

This gives you a tree structure of directories/files whose total size is ≥ 1 GB.

---

✅ If you want only the largest offenders (say top 20 big files/directories over 1 GB), you can do:

find /path/to/search -type f -size +1G -exec du -h {} + | sort -hr | head -20

---

For this, I had to sudo apt install tree

started using github projects for task tracking

How to use github projects - a getting started video - 
https://www.youtube.com/watch?v=oPQgFxHcjAw 

Opened a private repo with issues, and with ChatGPT's help, linked it to a project. (Projects --> New Project in the github web UI for that repo). Automation to create a new issue worked after some tweaking - the github actions yml is copy-pasted below:

name: Instantiate Planetarium Show Task
on:
  workflow_dispatch:
    inputs:
      show_title:
        description: 'Title of the Planetarium Show'
        required: true
      deadline:
        description: 'Deadline'
        required: false

permissions:
  issues: write
  contents: read
  
jobs:
  create_task_issue:
    runs-on: ubuntu-latest
    steps:
      - name: Build issue content
        run: |
          echo "## Show Details" > issue.md
          echo " " >> issue.md
          echo "**Show Title:**  " >> issue.md
          echo "${{ github.event.inputs.show_title }}" >> issue.md
          echo "" >> issue.md
          echo "**Deadline:**  " >> issue.md
          echo "${{ github.event.inputs.deadline }}" >> issue.md
          echo "" >> issue.md
          echo "## Steps" >> issue.md
          echo "" >> issue.md
          echo "- [ ] Create script" >> issue.md
          echo "- [ ] Create Hindi translation" >> issue.md
          echo "- [ ] Create Telugu translation" >> issue.md
          echo "- [ ] Review Hindi translation" >> issue.md
          echo "- [ ] Review Telugu translation" >> issue.md
          echo "- [ ] Record Hindi VO" >> issue.md
          echo "- [ ] Record Telugu VO" >> issue.md
          echo "- [ ] Edit Hindi VO" >> issue.md
          echo "- [ ] Edit Telugu VO" >> issue.md
          echo "- [ ] Warp ready for our theatre" >> issue.md
          echo "- [ ] Eng-Hin-Tel audio" >> issue.md
          echo "- [ ] Final audio mix" >> issue.md
          echo "- [ ] Save backups" >> issue.md
          echo "- [ ] Post on website" >> issue.md
          echo "" >> issue.md
          echo "---" >> issue.md
          echo "" >> issue.md
          echo "" >> issue.md

        
      - name: Create issue from template
        uses: peter-evans/create-issue-from-file@v4
        with:
          title: "[Show] ${{ github.event.inputs.show_title }}"
          content-filepath: issue.md
          token: ${{ secrets.GITHUB_TOKEN }}
          labels: show,task

This script had to be re-written due to ChatGPT repeatedly using heredoc syntax which would clash with github's yml. I had to explicitly ask it to "Please use echo >> on each line, and no heredoc."

One more automation script to automatically move the items in the project from "TODO" to "In Progress" when any of the checkboxes above were clicked, failed. ChatGPT usually fails when the documentation is not complete and / or there are conflicting earlier documentation pages, as in this case. It would repeatedly suggest using

permissions:
  issues: write
  projects: write

when actually such a permission is no longer accepted by Github - it was part of "classic projects" - https://docs.github.com/en/enterprise-server@3.16/issues/planning-and-tracking-with-projects/creating-projects/migrating-from-projects-classic. 

Apparently with projectsV2, the REST api does not work, only GraphQL API works. And it needs a classic personal access token to work. And the code generated by ChatGPT had errors which I could not get it to fix after multiple iterations over multiple days, so I just left updates to be done manually by clicking on the drop-down seen in the linked project for each issue. I could probably read the GraphQL API documentation and try again, but right now, not motivated to do so.

unable to use key-based auth

One of our servers had been upgraded, and after the upgrade, even though my public key was in the authorized_keys file, I was unable to ssh into that machine using a script with key-based authentication. 

It turned out the script which I was using to connect had the parameter

-o PubkeyAcceptedKeyTypes=ssh-rsa

which was necessary for the earlier server - earlier post about this - but not necessary for the current server - removing this parameter allowed the script to log in using key-based auth.

forcing 2FA for all cloudflare users

 After a cloudflare advisory asking us to rotate credentials due to a "Compromise Impacting Cloudflare Salesforce Cases", I was asked to force 2FA for all cloudflare users. 

https://developers.cloudflare.com/fundamentals/user-profiles/2fa/

The force-2FA option is available for super-admins in the cloudflare admin panel, at
Manage Account > Members > Members 2FA enforcement.  

getting referral ids for Moodle course enrolments

There was a request to store and report the referral ids passed via GET request from another site onto one of our Moodle instances. I turned to ChatGPT (free) for help. 

For the immediate need of getting referral ids from past logs, it suggested some python code. Which, of course, did not work on the first try. But 5 iterations later, got something useful out of it, which is shared at
https://github.com/hn-88/moodle-plugin-local-referrals/tree/main/temporary%20python%20code 

Then, it suggested writing a local plugin, and also suggested skeleton code. But the suggested code would not work if the user was not logged in to our site before clicking the referral link on the external site. It suggested saving the referral code to $SESSION. But that did not seem to work. Debugging to the Apache error log using lines like 
error_log('[local_referrals] function on_user_enrolled() ... ');
the referral id in $SESSION was being set to null when the user was logging in to our site.

Tried asking gemini.google.com (free) since ChatGPT seemed to be caught in a loop and unable to provide a fix. Gemini immediately looked up Moodle documentation and said that clearing the session (starting a new session) when a user logs on to Moodle is a known feature - and so we should store the referral id elsewhere - either in the database or in a cookie. Gemini in "Deep Research" mode also wrote a couple of essays about the issue, which could be summarized as my earlier sentence!

Then Gemini generated some code, but the code was incomplete. Asked ChatGPT to complete it - then asked Gemini to correct the errors introduced by ChatGPT - and the result is at
https://github.com/hn-88/moodle-plugin-local-referrals/ 

Some excerpts from the chats with ChatGPT and Gemini are in the wiki of that github repo - 
https://github.com/hn-88/moodle-plugin-local-referrals/wiki

Microsoft Azure - no meaningful way to contact regarding technical issues for non-profits

We're running some VMs on Microsoft Azure with free credits provided to non-profits. This means that in the Azure support section on the Azure portal, only billing support is visible, not technical support - that would probably need a paid plan. 

There was an intermittent network issue with one of our VMs - Ubuntu update not being able to reach the servers, ssh into the VM failing intermittently, and so on. 

I changed the type of support from Technical to Subscription, and then was able to submit the report.

Some more data points for troubleshooting -

1. Our other Azure VMs in other regions do not have this issue.

2. Some of our team members had issues connecting to this server on Aug 19 2025, between 3 & 7 pm IST. At around 8:45 pm IST, I could log on to the server without issues, I used BSNL as my ISP. On Aug 20, 11:30 am, the team at Hyderabad could not connect to the VM using Jiofiber ISP, but I could connect via Telexair located in Bangalore.

3. Rebooting the VM itself did not help - probably some other routing issue at your Central India location.

Edit 25 Sep - But there was no response from Microsoft other than that they would look into it - but the problem seemed to have been resolved in a few hours. Twenty days later, there was an email from the VM team, saying that the network team had passed on the issue to the VM team, so I just replied, "We have not had issues after my last updated to the ticket. You can call me if you wish between 9 and 11.30 am Indian Standard time or between 12.30 and 3.30 pm IST, but I have nothing further to add."

Google's lack of human support is well known. So, Azure is more of the same. I believe this is the norm for all these "cloud" companies.

resizing an Azure VM

Just going to portal.azure.com > VM > Overview > Sizing, choosing a different size, and hitting Resize did not work as planned, though a message popped up saying the VM would be restarted on resizing. After five minutes, stopped the VM - then it resized - then started it up again. MySQL took a couple of minutes to come up, so initially the Moodle instances complained about database connection failure.

admin takeover of subdomain in Microsoft Azure

 One of the subdomains of one of our domains had users actively using OneDrive and so on, and the admin of that domain wanted to implement some policies / restrict some users / or something like that. Looking around for options,

https://learn.microsoft.com/en-us/entra/identity/users/domains-admin-takeover

Probably need to do internal takeover since OneDrive etc don't support external takeover. But right now, nothing has been done.

connections problems to an AWS server

We were using ssh to connect to an AWS VM, and the developers who would log in from Hyderabad reported that they were unable to ssh into it. I too had problems connecting initially. 

Checking the AWS console to see if the IP address had changed, found that the IP, port etc were unchanged. Later, I was able to connect with ssh. 

So, maybe it was an internet connection issue? Or a routing issue with AWS or with some of the ISPs?

using ABeeZee TTF font on a Moodle instance

There was a request to enable the ABeeZee TTF font on one of our Moodle instances, since it is recommended for inclusivity etc. https://fonts.google.com/specimen/ABeeZee/about says it is a "children's learning font".

It was later clarified that this font was desired for all content on the site and for the custom certificates. So,

1. For adding custom fonts to course content and body text, I believe the method is to edit the CSS of the current theme, as mentioned here - 

https://docs.moodle.org/500/en/How_to_add_custom_fonts_in_a_theme

In our case, the theme is Moove.

2. For the custom certificate plugin, the method followed - adding the font using "Manage PDF fonts" using the TCPDF plugin, is the method we have followed earlier. 

https://hnsws.blogspot.com/2023/10/request-for-new-font-in-moodle-instance.html

So it should be working for creating custom certificates. 

For changing the font in the theme, it is relatively easy to implement if we were to change all content to use this font.

Site administration → Appearance → Additional HTML

and I've added inside the "Within Head" box

<link rel="stylesheet" href="https://fonts.googleapis.com/css2?family=ABeeZee&display=swap">

and inside the Moodle theme (which is used on our sites) advanced settings box, (which has the url oursite.org/admin/settings.php?section=themesettingmoove#theme_moove_advanced )

body {

    font-family: 'ABeeZee', sans-serif !important;

}

(which makes the entire body of the html page into ABeeZee.)

This was implemented for two of our sites after first trying it on our development instance.

making persistent USB Ubuntu 24.04

https://itsfoss.com/ubuntu-persistent-live-usb/

doing the mkusb method since Windows would not recognize the drive which had a bootable windows setup disk on it?

sudo add-apt-repository ppa:mkusb/ppa

sudo apt update

sudo apt install --install-recommends mkusb mkusb-nox usb-pack-efi

Then mkusb - UI-based tool - is available on the system menu.

Virtualbox error - Cannot register - the hard disk already exists

Via https://stackoverflow.com/questions/44114854/virtualbox-cannot-register-the-hard-disk-already-exists

For me, the menu path in Virtualbox 7.1 was:

Tools -> (Hamburger menu) -> Media -> (choose the one to remove) -> Remove button on toolbar

turn off gmail "smart" summaries

 via https://www.digitalinformationworld.com/2025/05/gmail-is-reading-and-summarizing-your.html

On Android --> GMail app --> Hamburger menu --> Settings --> choose the account for which you want to disable --> Untick smart features checkbox

lots of github and Mac learnings

Trying to compile OpenSpace on MacOS - 

https://github.com/hn-88/OpenSpace/tree/ToCompileOnXCode16.4

https://github.com/hn-88/OpenSpace-AppleSiliconMac/actions

https://github.com/hn-88/OpenSpace-IntelMac/actions

Learnt a lot of things from Copilot and ChatGPT - since I didn't have any humans available to ask specific questions. Also bumped against the limitations of the LLMs - sometimes they give wrong advice with full confidence! Perhaps adding prompts to cite sources, mention confidence level etc might help. Some points, in no particular order.

•  XCode support for some C++ features is missing - C++ language support - and in some cases, AppleClang's interpretation is stricter than gnu c++ or MSVC. Also, some flags specific to MacOS would change from old versions to new versions of MacOS. And homebrew support for older Macs is spotty. So, supporting Macs is a challenge.
  
  
• Verifying and testing on github runners has some issues due to their headless nature.
  
  
• Uploading to draft releases works best if that draft release is pre-created - sometimes the tag creation takes a few seconds, leading to failure in uploading the artifact to the draft release. Or, can create a new tag, wait for a few seconds, and then upload the artifact to draft release using a personal access token - PAT.  
  
  
• MacOS, especially after 15.2, does not make it easy to run unsigned apps. Some workarounds which exist now, like removing the quarantine flag from the internal binary, may not work in the future. It is possible that Debug builds may work and (unsigned) Release builds may not work as in the case of Android apk files - unless we add a codesign step after the build.
  
  
• Even with the bundling of dylibs using otool, the binary still seems to look for dylibs in /opt/homebrew etc. So, the bundling code suggested by ChatGPT still has flaws. Or maybe there are some dependency calls which are only at runtime and hence not detected by otool. 

2FA required for Google Workspace admin users

Passed on Google's message that two-factor authentication / 2FA / 2-Step Verification (2SV) is required for Google Workspace admins starting August 13 2025 - 

.... please alert any of the users who have been given access to admin.google.com in case they have not enabled 2FA already.

How to enable 2FA:

Start by logging on to your Google account for our.domain.org

• Click your avatar (top-right) and choose Manage your Google Account.
• Click Security in the left pane.
• In the Signing in to Google section, click 2-Step Verification.
• Click the Get Started button.
• Enter your password (if requested).
• Click Continue.
• Click Send to send a text message to your phone.
• Enter the 6 digit code sent to your phone and click Next.
• Click Turn On to finish the process.
• Your account is now protected with 2 step verification.

manual time sync on Ubuntu 24.04

Asked ChatGPT, and it gave the long answer below. The short answer is:

sudo systemctl restart systemd-timesyncd

Long answer pasted below:

To manually trigger an NTP (Network Time Protocol) time update on Ubuntu, you can use the following steps depending on whether you're using systemd-timesyncd, chrony, or ntpd.

🔍 Step 1: Check which time sync service is in use

Run:

timedatectl show | grep -i ntp

Also try:

systemctl status systemd-timesyncd

systemctl status chronyd

systemctl status ntp

✅ If you're using systemd-timesyncd (default on newer Ubuntu versions):

Trigger a manual update:

sudo systemctl restart systemd-timesyncd

Optionally, you can force immediate sync using timedatectl:

timedatectl timesync-status

To see the current status or wait a bit and verify with:

timedatectl status

⚠️ Note: systemd-timesyncd does not support manual "force now" updates — it syncs on its own schedule after restarting.

✅ If you're using chrony:

Force an update:

sudo chronyc makestep

This forces an immediate correction if the clock is off.

You can verify with:

chronyc tracking

✅ If you're using ntpd:

Force an update:

sudo ntpd -gq

-g allows a large initial offset

-q exits after setting the time

Then restart the daemon:

sudo systemctl restart ntp

🕒 Verify time sync

timedatectl

Look for: System clock synchronized: yes

solution for rendering black frame & moving a Blender keyframe to a different time

The blend file linked from

https://www.youtube.com/watch?v=Jh1p82AJkaU

would not render out of the box on Cycles or Eevee with Blender 4. (the tutorial was made for Blender 2.8). Thought that one possible reason was that I had modified the number of render samples:

"Samples should not be lower than volumetric samples."

But Cycles was rendering with multiple samples, up to the final moment, when the render would go black. ChatGPT confirmed my hunch that this was due to some compositing settings. Disabled compositing in Render properties --> Post-processing - and then the render was showing up. 4K renders of each frame was taking around 5 to 10 minutes on the 1060TX GPU. Sheepit render farm rendered the project in less than a day, after making modifications to the length of the animation and using a 180 degree fisheye "fulldome" panoramic camera.

For moving keyframes - to make the animation 10x longer, move a keyframe at frame 25 to frame 250, as an example - https://blender.stackexchange.com/questions/221434/how-to-move-selected-keyframes-to-current-frame-or-a-specified-frame-number#221446

"Hovering on the action editor or dopesheet you can select the keyframes you want to move, set the time cursor to the desired frame and press S,X,0,Enter."

The final product is below. Fulldome (180 degree fisheye) 4K video suitable for planetariums. The full 4K resolution video is made available at archive.org. 

just git push after git checkout -b existingbranch

If we check out an existing branch, make modifications and want to push to that branch, we just have to do 

git push

and not 

git push -b existingbranch

or else we would get the error

existingbranch already exists

https://stackoverflow.com/questions/53046763/fatal-a-branch-named-new-branch-already-exists

https://www.geeksforgeeks.org/git/default-behaviour-of-git-push-without-a-branch-specified/

descreen and perspective correction with gimp

To remove moire patterns due to silk-screened photos from prints, gmic plugin's fourier filter can be used. To install - gimp-gmic is in software manager of Linux Mint.

https://www.youtube.com/watch?v=3137dDa6P4s

According to this video, the way to do it -
1. first fourier transform in gmic
2. Then make the white dots black for everything except centre
3. then inverse
4. then smooth medium (also in gmic)

(This page refers to a specific descreen plugin, but that is for Gimp 2.0 and the link also seems broken, so didn't try out that.)

For perspective correction, especially for "re-clicked" photos taken with mobile from an angle,

Tools --> Transform tools --> Perspective --> choose Corrective (backward) for Direction, 

put the four corners in the four corners of the photo to be corrected. Then, the four corners will be stretched to the aspect ratio of the original image. After that, some Transform --> Scale changes may be needed, with aspect ratio unlocked, in case the photo being corrected has a different aspect ratio. 

http working but not https

There was a report that one of our websites was down. Reported to PB, his reply was:

Found that http was working but not https

Then did the following:

root@ourserver:openssl s_client -connect our.domain.org:443 -servername our.domain.org

and got the error message

verify error:num=20:unable to get local issuer certificate

# comment out in  /etc/apache2/sites-enabled/our.domain.org-le-ssl.conf

# SSLCertificateChainFile /etc/letsencrypt/live/our.domain.org/chain.pem

and then restarted apache2