Issues compiling FFmpeg

[RonaldGalea]

Hello!

I'm encountering some issues compiling FFmpeg following the guide

Specifically, I see the following errors (measure_time is just a small script to print the runtime):

root@29b4c3187cd6:/app/gg/compilation_benchmarks/ffmpeg/experiment/build# ./measure_time gg force --jobs 64 --engine lambda ffmpeg
→ Loading the thunks...  done (313 ms).
↗ Uploading 6197 files (138.8 MiB)... done (8454 ms).
/__gg__/ar: unable to copy file 'libavcodec/libavcodec.a'; reason: No error information
rmdir /tmp/thunk-execute.MJUoGP: Directory not empty
std::exception
 `T45XGjJ13b62vhKGXQ8zbaG6MhFJF5rgRusUOk4vcbFs0001c1cb': process exited with failure status 1

gg-force: execution failed: T45XGjJ13b62vhKGXQ8zbaG6MhFJF5rgRusUOk4vcbFs0001c1cb
Execution time: 44787 milliseconds

Sometimes, it fails in different ways (here I killed it because it was hanging):

root@29b4c3187cd6:/app/gg/compilation_benchmarks/ffmpeg/experiment/build# ./measure_time gg force --jobs 64 --engine lambda ffmpeg
→ Loading the thunks...  done (309 ms).
↗ Uploading 6197 files (138.8 MiB)... done (7938 ms).
invalid response: {"returnCode": -6, "stdout": "terminate called after throwing an instance of 'ssl_error'\n  what():  SSL_read: error:00000005:lib(0):func(0):DH lib\n"}
[warning] operational failure: T8M8L.j4tRIUGSnJXoMp0_gmZ0bfRNmhcyGWqADtFciw000037a0
^Cgg-force: killed by signal
Execution time: 140805 milliseconds

I tried various ways of fixing the problem (deleting gg caches, resetting the entire build environment), but issues like the above still persist. Of course, just running make locally works fine.

Relevant information about the environment:

• Ubuntu 20.04.6 LTS
• gcc (Ubuntu 7.5.0-6ubuntu2) 7.5.0
• g++ (Ubuntu 7.5.0-6ubuntu2) 7.5.0

As a side note, I noticed gg infer takes around 30 seconds, is this expected or might I have an issue with my setup? I have an AMD Ryzen 5 3600XT CPU (12 hyperthreads) and am running gg infer make -j$(nproc)

As a second note, I had to change the Lambda function runtime from python 3.6 to 3.9, since 3.6 is no longer supported, could this be related to the issues I'm seeing?

Edit: For completeness, I also increased the hard-coded limit for disk size, but modified the function to have enough storage accordingly. If this was the issue, I would be seeing a disk space related error, right?

If necessary, I can provide a Dockerfile along with full scripts and steps to reproduce. Thank you in advance!