[CI] Add backward ABI-compatibility testing to pre-commit #19719

aelovikov-intel · 2025-08-05T22:04:50Z

No description provided.

.github/workflows/sycl-linux-precommit.yml

#19719 broke `workflow_dispatch` mode for manual E2E task invocation for compatibility testing as we don't have the required `binaries_artifact` input in that mode. Github limits number of input parameters in `workflow_dispatch` mode, so restore the functionality without introducing extra input.

#19744) #19719 broke `workflow_dispatch` mode for manual E2E task invocation for compatibility testing as we don't have the required `binaries_artifact` input in that mode. Github limits number of input parameters in `workflow_dispatch` mode, so restore the functionality without introducing extra input.

aelovikov-intel · 2025-08-11T14:11:51Z

devops/compat_ci_exclude.sycl-rel-6_2

+
+# Need more investigation by the author:
+
+# https://github.com/intel/llvm/pull/18314


@Alexandr-Konovalov , looks like your #18314 broke backward compatibility here, see
https://github.com/intel/llvm/actions/runs/16882134385 (using nightly-2025-05-08, pass)
https://github.com/intel/llvm/actions/runs/16882361500 (using nightly-2025-05-09, fail)
Please investigate and either fix (might need backporting to sycl-rel-6_3) or create a PR updating this file describing why it wasn't an ABI-break or was allowed/desired/approved.

Thank you, an interesting issue! Opened [SYCL] Add copy/move ctors and assignment operators to sycl::detail::optional #19775.

aelovikov-intel · 2025-08-11T14:23:26Z

devops/compat_ci_exclude.sycl-rel-6_2

+Assert/assert_in_simultaneously_multiple_tus.cpp
+Assert/assert_in_simultaneously_multiple_tus_one_ndebug.cpp
+
+# https://github.com/intel/llvm/pull/17442


@steffenlarsen , looks like your #17442 broke backward compatibility here, see
https://github.com/intel/llvm/actions/runs/16882650449 (using nightly-2025-05-01, pass)
https://github.com/intel/llvm/actions/runs/16882716814 (using nightly-2025-05-02, fail)
Please investigate and either fix (might need backporting to sycl-rel-6_3) or create a PR updating this file describing why it wasn't an ABI-break or was allowed/desired/approved.

Will do! Great addition, by the way!

See #19764.

aelovikov-intel · 2025-08-11T14:52:13Z

devops/compat_ci_exclude.sycl-rel-6_2

+NewOffloadDriver/split-per-source-main.cpp
+NewOffloadDriver/sycl-external-with-optional-features.cpp
+
+# https://github.com/intel/llvm/pull/18277


@igchor , looks like your #18277 broke backward compatibility here, see

https://github.com/intel/llvm/actions/runs/16883391382 (using nightly-2025-05-19, pass)
https://github.com/intel/llvm/actions/runs/16883503382 (using nightly-2025-05-23, fail)
Please investigate and either fix (might need backporting to sycl-rel-6_3) or create a PR updating this file describing why it wasn't an ABI-break or was allowed/desired/approved.

aelovikov-intel · 2025-08-11T14:57:58Z

devops/compat_ci_exclude.sycl-rel-6_2

+# https://github.com/intel/llvm/pull/19328
+Adapters/interop-level-zero-buffer-ownership.cpp


@igchor , looks like your #19328 broke backward compatibility here, see

https://github.com/intel/llvm/actions/runs/16883683175 (using nightly-2025-07-10, pass)
https://github.com/intel/llvm/actions/runs/16883690547 (using nightly-2025-07-11, fail)
Please investigate and either fix (might need backporting to sycl-rel-6_3) or create a PR updating this file describing why it wasn't an ABI-break or was allowed/desired/approved.

aelovikov-intel · 2025-08-11T15:10:00Z

devops/compat_ci_exclude.sycl-rel-6_2

+# https://github.com/intel/llvm/pull/18403 (pulldown, so probably offload-tools
+# team should be looking into this)
+DeviceImageDependencies/NewOffloadDriver/dynamic.cpp
+DeviceImageDependencies/NewOffloadDriver/free_function_kernels.cpp
+DeviceImageDependencies/NewOffloadDriver/math_device_lib.cpp
+DeviceImageDependencies/NewOffloadDriver/objects.cpp
+DeviceImageDependencies/NewOffloadDriver/singleDynamicLibrary.cpp
+NewOffloadDriver/aot-gpu.cpp
+NewOffloadDriver/buffer.cpp
+NewOffloadDriver/multisource.cpp
+NewOffloadDriver/spirv_device_obj_smoke.cpp
+NewOffloadDriver/split-per-source-main.cpp
+NewOffloadDriver/sycl-external-with-optional-features.cpp


@intel/dpcpp-tools-reviewers , looks like #18403 pulldown broke backward compatibility here

https://github.com/intel/llvm/actions/runs/16883874414 (using nightly-2025-05-16, pass)
https://github.com/intel/llvm/actions/runs/16883880550 (using nightly-2025-05-17, fail)
Please investigate and either fix (might need backporting to sycl-rel-6_3) or create a PR updating this file describing why it wasn't an ABI-break.

It was suggested to me to tag @asudarsa here.

aelovikov-intel · 2025-08-11T15:18:57Z

devops/compat_ci_exclude.sycl-rel-6_2

+# https://github.com/intel/llvm/pull/18565
+# This one should have probably be done in multiple PRs, first improving the
+# test's CHECKs and then making an actual functional change.
+#
+# Based on the title, I'd expect to see reduction in number of ur*retain/release
+# calls, but it only moves a few ur*release CHECKs, so I'll let the author
+# update this explanation.


@Alexandr-Konovalov , please upload a PR that updates this comment with explanation how the change wasn't ABI-breaking.

Opened [CI] Add clarification for KernelAndProgram/disable-caching change #19822

aelovikov-intel · 2025-08-11T15:20:30Z

devops/compat_ci_exclude.sycl-rel-6_2

+# binaries built with sanitizers?
+Sanitizer
+
+# https://github.com/intel/llvm/pull/19238


@AndreiZibrov , please upload a PR that updates this with an explanation why this was desired (I'd expect something along the lines of allowing ABI breaks in an experimental extension).

Also tagging @AlexeySachkov

@aelovikov-intel - Do you have the output of this failure at hand?

Easy to generate :)
https://github.com/intel/llvm/actions/runs/16911593954

Thanks! Clarification here: #19774

aelovikov-intel · 2025-08-11T15:36:32Z

devops/compat_ci_exclude.sycl-rel-6_2

+# https://github.com/intel/llvm/pull/17955, experimental extension
+AsyncAlloc/device/async_alloc_from_pool.cpp
+AsyncAlloc/device/async_alloc_zero_init.cpp
+AsyncAlloc/device/ooo_queue_async_alloc_from_pool.cpp


@intel/sycl-graphs-reviewers , can you please check if that break was intentional and if so update the comment here?

Using 2025-06-18 (pass): https://github.com/intel/llvm/actions/runs/16884717829
Using 2025-06-19 (fail): https://github.com/intel/llvm/actions/runs/16884725131

intel#19744) intel#19719 broke `workflow_dispatch` mode for manual E2E task invocation for compatibility testing as we don't have the required `binaries_artifact` input in that mode. Github limits number of input parameters in `workflow_dispatch` mode, so restore the functionality without introducing extra input.

xtian-github · 2025-08-14T17:56:48Z

@cdai2 Please take a look this PR to see what is impact on Sanitizer compatibility.

intel#19719 and intel#19761 added pre-commit jobs to run E2E tests pre-built with latest "open-source" releases against the newly built sycl-toolchain libraries. Those can fail if either an actual break is happenning or if the test was doing some `FileCheck`ing and that output has changed in some way (which might not be an actual ABI break). However, I think the testing is still good enough to require an explicit approvals by folks in charge of ABI breaking changes. For the case of just output change the author should be able to convince owners that the change isn't ABI-breaking relatively easily.

…19820) #19719 and #19761 added pre-commit jobs to run E2E tests pre-built with latest "open-source" releases against the newly built sycl-toolchain libraries. Those can fail if either an actual break is happenning or if the test was doing some `FileCheck`ing and that output has changed in some way (which might not be an actual ABI break). However, I think the testing is still good enough to require an explicit approvals by folks in charge of ABI breaking changes. For the case of just output change the author should be able to convince owners that the change isn't ABI-breaking relatively easily.

aelovikov-intel force-pushed the ci-abi-compat branch 4 times, most recently from 30a6942 to 97b4425 Compare August 6, 2025 15:11

aelovikov-intel had a problem deploying to WindowsCILock August 6, 2025 15:11 — with GitHub Actions Error

aelovikov-intel force-pushed the ci-abi-compat branch from 97b4425 to 4315d9a Compare August 6, 2025 15:29

aelovikov-intel had a problem deploying to WindowsCILock August 6, 2025 15:29 — with GitHub Actions Error

aelovikov-intel force-pushed the ci-abi-compat branch from 4315d9a to d1fe046 Compare August 6, 2025 15:31

aelovikov-intel had a problem deploying to WindowsCILock August 6, 2025 15:31 — with GitHub Actions Error

aelovikov-intel force-pushed the ci-abi-compat branch from d1fe046 to 3ef89a0 Compare August 6, 2025 15:47

aelovikov-intel temporarily deployed to WindowsCILock August 6, 2025 15:47 — with GitHub Actions Inactive

aelovikov-intel had a problem deploying to WindowsCILock August 6, 2025 16:12 — with GitHub Actions Error

aelovikov-intel force-pushed the ci-abi-compat branch from 3ef89a0 to 0642488 Compare August 6, 2025 16:14

aelovikov-intel temporarily deployed to WindowsCILock August 6, 2025 16:14 — with GitHub Actions Inactive

aelovikov-intel had a problem deploying to WindowsCILock August 6, 2025 16:35 — with GitHub Actions Error

aelovikov-intel temporarily deployed to WindowsCILock August 6, 2025 16:35 — with GitHub Actions Inactive

aelovikov-intel force-pushed the ci-abi-compat branch from 0642488 to a1bdc90 Compare August 6, 2025 17:59

aelovikov-intel temporarily deployed to WindowsCILock August 6, 2025 18:00 — with GitHub Actions Inactive

aelovikov-intel temporarily deployed to WindowsCILock August 6, 2025 18:35 — with GitHub Actions Inactive

aelovikov-intel force-pushed the ci-abi-compat branch from a1bdc90 to e21ec5d Compare August 6, 2025 18:53

aelovikov-intel temporarily deployed to WindowsCILock August 6, 2025 18:54 — with GitHub Actions Inactive

aelovikov-intel temporarily deployed to WindowsCILock August 6, 2025 19:19 — with GitHub Actions Inactive

aelovikov-intel force-pushed the ci-abi-compat branch from e21ec5d to be22afc Compare August 6, 2025 21:12

aelovikov-intel temporarily deployed to WindowsCILock August 6, 2025 21:12 — with GitHub Actions Inactive

aelovikov-intel had a problem deploying to WindowsCILock August 6, 2025 21:33 — with GitHub Actions Error

aelovikov-intel force-pushed the ci-abi-compat branch from be22afc to 3c00aee Compare August 6, 2025 21:34

Use default image_options

fd6a97d

aelovikov-intel commented Aug 7, 2025

View reviewed changes

.github/workflows/sycl-linux-precommit.yml Show resolved Hide resolved

aelovikov-intel temporarily deployed to WindowsCILock August 7, 2025 16:46 — with GitHub Actions Inactive

aelovikov-intel temporarily deployed to WindowsCILock August 7, 2025 17:06 — with GitHub Actions Inactive

aelovikov-intel temporarily deployed to WindowsCILock August 7, 2025 17:32 — with GitHub Actions Inactive

aelovikov-intel requested a review from uditagarwal97 August 7, 2025 18:00

uditagarwal97 approved these changes Aug 7, 2025

View reviewed changes

aelovikov-intel merged commit efe036e into intel:sycl Aug 7, 2025
36 of 37 checks passed

aelovikov-intel deleted the ci-abi-compat branch August 7, 2025 18:30

aelovikov-intel mentioned this pull request Aug 7, 2025

[CI] Use E2E container binaries in run-only manual E2E task dispatch #19744

Merged

aelovikov-intel commented Aug 11, 2025

View reviewed changes

ggojska pushed a commit to ggojska/llvm that referenced this pull request Aug 12, 2025

[CI] Add backward ABI-compatibility testing to pre-commit (intel#19719)

3a4271e

aelovikov-intel mentioned this pull request Aug 18, 2025

[DevSAN] Change DeviceType to specialization constant #19798

Open

aelovikov-intel mentioned this pull request Aug 18, 2025

[CI] Change codeowners for backward ABI-compatibility exclude lists #19820

Merged


		# Need more investigation by the author:

		# https://github.com/intel/llvm/pull/18314

		# https://github.com/intel/llvm/pull/19328
		Adapters/interop-level-zero-buffer-ownership.cpp

[CI] Add backward ABI-compatibility testing to pre-commit #19719

[CI] Add backward ABI-compatibility testing to pre-commit #19719

Uh oh!

Conversation

aelovikov-intel commented Aug 5, 2025

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

steffenlarsen Aug 12, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

xtian-github commented Aug 14, 2025

Uh oh!

Uh oh!

steffenlarsen Aug 12, 2025 •

edited

Loading