Skip to content

Update workflows for public repo - #1

Merged
qti-jkilpatrick merged 1 commit into
mainfrom
dev/jkilpat/upgrade-artifact-actions
Jan 28, 2026
Merged

qti-jkilpatrick merged 1 commit into
mainfrom
dev/jkilpat/upgrade-artifact-actions

Conversation

@qti-jkilpatrick

@qti-jkilpatrick qti-jkilpatrick commented Jan 26, 2026 •

Copy link
Copy Markdown
Collaborator
  • Upgrade upload/download-artifact versions.
  • Temporarily disable PR workflow jobs requiring an ARM64 Windows device.

@qti-jkilpatrick
qti-jkilpatrick force-pushed the dev/jkilpat/upgrade-artifact-actions branch 4 times, most recently from ad77fd3 to 97da0a4 Compare January 27, 2026 17:41
* Upgrade upload/download-artifact versions
@qti-jkilpatrick
qti-jkilpatrick force-pushed the dev/jkilpat/upgrade-artifact-actions branch from 97da0a4 to a54f2df Compare January 27, 2026 18:45
@qti-jkilpatrick
qti-jkilpatrick merged commit 669158b into main Jan 28, 2026
20 checks passed
@qti-jkilpatrick
qti-jkilpatrick deleted the dev/jkilpat/upgrade-artifact-actions branch January 28, 2026 00:43
@qti-jkilpatrick
qti-jkilpatrick restored the dev/jkilpat/upgrade-artifact-actions branch January 28, 2026 17:54
@qti-jkilpatrick
qti-jkilpatrick deleted the dev/jkilpat/upgrade-artifact-actions branch January 28, 2026 19:19
qti-kromero added a commit that referenced this pull request Feb 17, 2026
* Fix bug where wheel name was not getting wheel name suffix added to it
* Undoes some of storage compromises introduced by #1 and #14
kuanyul-qti pushed a commit that referenced this pull request Feb 23, 2026
* Fix bug where wheel name was not getting wheel name suffix added to it
* Undoes some of storage compromises introduced by #1 and #14
@chunghow-qti chunghow-qti mentioned this pull request May 19, 2026
3 tasks done
chunghow-qti added a commit that referenced this pull request Jun 3, 2026
QnnCPUBackendTests.UDO_Op_MyAdd / QnnHTPBackendTests.UDO_Op_MyAdd dlopen
a generated libMyAddOpPackage_*.so. Allocations inside that package's
constructor (call_init -> lib _init) and at runtime resolve to anonymous
mmap regions, so LSan reports them as:

  Direct leak of N byte(s) in M object(s) allocated from:
      #0 operator new
      #1 0x...  (<unknown module>)
      #2 call_init elf/dl-init.c:70

This is not an ORT-side leak, but two of our existing safeguards both
miss it:
  - tools/ci_build/lsan_suppressions.txt's "leak:libQnnHtp.so" needs a
    module path on the frame; <unknown module> carries none.
  - asan_filter_leaks.sh assumes "Direct leak == ORT/test code" and
    fails CI.

Add "--gtest_filter=-*UDO_Op_MyAdd*" so the rest of the suite runs
strictly under ASan. Re-enable once the UDO test tears down its op
package on shutdown or asan_filter_leaks.sh learns to recognize
<unknown module> roots as third-party noise.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants