Make the DOM codes reproducible #12679
Conversation
|
Thank you for submitting a pull request! If this is your first PR, make sure to add yourself to AUTHORS. |
Whilst working on the Reproducible Builds effort [0] I noticed that emscripten could not be built reproducibly. This is because it generates a switch/case statement using Python's random.randint/random.uniform functionality. This patch seeds this random component using the SOURCE_DATE_EPOCH environment variable [1] so that, iff this variable is present, it will deterministically generate the same output. I originally filed this bug in Debian as #973595 [2]. [0] https://reproducible-builds.org/ [1] https://reproducible-builds.org/specs/source-date-epoch/ [2] https://bugs.debian.org/973595
|
But that script is not used during any builds, as far I can tell. It looks like that scripts only job is to create So for sure any use of random in this script is not effecting any builds. Given that we seem to bascially never use this script I think we should probably consider removing it anyway. |
Interesting, as that script is definitely being run during the Debian build. |
Can you point me to where that is? |
|
I do agree this script should be deterministic though (otherwise all the value will change each time it is run which is clearly not indeneded).. so perhaps a constant seed is better. |
|
Looks like that's here: https://sources.debian.org/src/emscripten/2.0.8~dfsg-6/debian/rules/#L171 (As for exactly why this was added, we would need to loop in the Debian maintainer for this particular package; we have reached the extent of my insight, alas.) |
|
Oh man.. that line should certainly be removed. I'm working on some other fixes to that file now. In the mean time, if you would like to land this change can you use a fixed seed? |
|
Sorry for any confusion but I'm not the Debian maintainer for this package so I am ultimately not in the position of being able to comment on whether that line can/should be removed - I merely found the unreproducibility and let you kind folks know. :) I would guess a fixed, hardcoded seed would work; indeed, this is in essence what my patch does in its own way. :) |
Embed the current hash seed values in the script so that running it repeatendly doesn't change the output. See #12679
Embed the current hash seed values in the script so that running it repeatendly doesn't change the output. See #12679
Embed the current hash seed values in the script so that running it repeatedly doesn't change the output. See #12679
|
I've landed two changes that should make this no longer a problem. If you feel like following up with downstream, that line should still be removed from the debian/rules file. |
Whilst working on the Reproducible Builds effort I noticed that emscripten could not be built reproducibly. This is because it generates a switch/case statement using Python's
random.randint/random.uniformfunctionality.This patch seeds this random component using the
SOURCE_DATE_EPOCHenvironment variable so that, iff this variable is present, it will deterministically generate the same output.(I originally filed this bug in Debian as #973595.)