test(profiling): with address sanitizer on bookworm with GH actions

[morrisonlevi]

Description

This adds bookworm images which were heavily based on the buster image:

• Many dependencies are newer and did not have to be built from scratch, which is nice.
• This doesn't run all the tests on bookworm, just builds and uploads images.
• However, only some images can be built. More work is required for PHP 7.3 and below to get working. I'm not sure how much more work is needed, but the two known issues are with intl and I think PHP 7.0 requires OpenSSL v1. Additionally, the shared extension versions have failures, something about ldap_connect having a mismatched signature, possibly: build fails with OpenLDAP 2.5.4 curl/curl#7004.

The key reason this PR was made is to add a PHP 8.3 NTS ASAN build, and then runs the profiling/tests/phpt tests with it. This is why I didn't continue working on the bookworm images so that all versions work; I really just need ASAN right now, and it wasn't working on Buster due to libasan5 vs libasan6 differences.

PROF-8286

Reviewer checklist

• Test coverage seems ok.
• Appropriate labels assigned.

[pr-commenter]

Benchmarks

Benchmark execution time: 2024-01-11 10:53:21

Comparing candidate commit a486091 in PR branch levi/bookworm with baseline commit 41cb273 in branch master.

Found 0 performance improvements and 2 performance regressions! Performance is the same for 16 metrics, 3 unstable metrics.

scenario:walk_stack/50

• 🟥 wall_time [+753.578ns; +757.005ns] or [+4.939%; +4.962%]

scenario:walk_stack/99

• 🟥 wall_time [+805.193ns; +809.729ns] or [+5.265%; +5.295%]

[bwoebi]

I do not understand the point of this file?

[morrisonlevi]

The current buster builds are "unversioned." This can cause issues (like the one you mention below with nts-asan failing). This file is a way to version the build, although it's not plumbed all the way through everything like the CI and such. We do this in the libdatadog images and it's worked nicely there.

[bwoebi]

I'm not sure whether parts of this are already pushed, but https://app.circleci.com/pipelines/github/DataDog/dd-trace-php/13970/workflows/b0aaa43d-7966-4958-a4c2-416367afda4c/jobs/3410571 looks somewhat related I suppose?

[morrisonlevi]

The push was from a different but related PR where I tried to do this on buster images. I have re-pushed images from that were built from master.

[morrisonlevi]

I've added lsof and rebuilt the relevant images. Once this is reviewed, I'll actually add back all the buster image sources which were deleted in this PR. Doing it in this way makes the diff much easier to review.

[realFlowControl]

The Cache build dependencies step did not cache anything due to:

Warning: Path Validation Error: Path(s) specified in the action for caching do(es) not exist, hence no cache is being saved.
Warning: Cache save failed.

See https://github.com/DataDog/dd-trace-php/actions/runs/7468310960/job/20347970685?pr=2432
I fixed the path in e34c5f6

It is still not working correctly, but this might be due to some race condition, as I started two actions at the same time (one through push and another through retry), so currently it looks like the cache key exists in a weird state.

We should have a look after merging in a new PR where we just cargo update, which will change the key and then we'll see.

So feel free to merge 🎉