GH-99108: Make vectorized versions of Blake2 available on x86, too

[msprotz]

Per #99108 (comment): trying a quick fix to see if this builds.

@gpshead can you buildbot this for Debian x86? Thanks.

If it works, I'll patch & fix upstream, then pull it in accordingly into CPython. Thank you!

• Issue: Replace built-in hashlib with verified implementations from HACL* #99108

[msprotz]

(The SBOM build will fail for the time being.)

[encukou]

!buildbot x86

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @encukou for commit 091fd62 🤖

The command will test the builders whose names match following regular expression: x86

The builders matched are:

• x86-64 MacOS Intel NoGIL PR
• x86-64 macOS PR
• x86 Debian Installed with X PR
• x86-64 MacOS Intel ASAN NoGIL PR
• x86 Debian Non-Debug with X PR

[msprotz]

Looks like this compiles but gives an unrelated error in tests. @gpshead @encukou do you agree? If so, I'll fix it upstream

[encukou]

Yup, I agree!
I'll need to fix that ctypes error now that it's uncovered :)

[vstinner]

buildbot/x86 Debian Installed with X PR — Build done.
buildbot/x86 Debian Non-Debug with X PR — Build done.

Tests failed on these two workers, but it's a known issue: #121938

This PR fix the build on these two workers: #125535

[vstinner]

LGTM

[vstinner]

(The SBOM build will fail for the time being.)

I ran make regen-sbom.

[vstinner]

If it works, I'll patch & fix upstream, then pull it in accordingly into CPython. Thank you!

Yeah, please fix upstream ;-)

[msprotz]

The fix just got merged a couple hours ago upstream, I need to run a build to refresh the correct folder, and then I'll push the refresh here (along with make regen-sbom). Thanks!

[vstinner]

@gpshead @encukou: Why did you block merging this PR? Do you prefer to wait until upstream is fixed, and then update Hacl in Python?

[encukou]

Yes.

[msprotz]

Ok this is pushed, let's see what CI says. @vstinner FYI, this would be an SBOM violation to push a local fix to an external vendored library, so this is why we have to propagate the fix upstream first... thanks!

[vstinner]

FYI, this would be an SBOM violation to push a local fix to an external vendored library

I didn't know that. We made many changes in mimalloc vendored copy.

[msprotz]

Maybe mimalloc gets a different treatment? The SBOM CI target fails if the vendored copy of HACL* is not exactly identical to upstream...

[gpshead]

For this one we set it up to be upstream-first and just track specific upstream versions. it works since we've got active maintainers. :)

[zware]

!buildbot x86 Deb

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @zware for commit acd04e6 🤖

The command will test the builders whose names match following regular expression: x86 Deb

The builders matched are:

• x86 Debian Non-Debug with X PR
• x86 Debian Installed with X PR

[zware]

This fixes the build issue on x86 Debian. Not sure what's up with the CI job checking the SBOM, though. I can't reproduce that failure locally (cc @sethmlarson for sbom insight).

[vstinner]

@msprotz: Can you re-run make regen-sbom?

[msprotz]

I thought I had run it. Let me run it again.

[msprotz]

I committed the suggested change, but make regen-sbom generates no diff on my machine still. Let's see if this fixes CI

[zware]

I could have left a clearer note, sorry. It turned out that the generated files check was failing on regen-sbom because Tools/build/generate_sbom.py behaves differently if CI is defined in the environment. In that case, it actually does the download and regenerates the checksum.

[msprotz]

Ah ok! Good to know for next time, thank you.