1796523 - GCC PGO build fails - source locations for function XY have changed, the profile data may be out of date

Status: NEW

(Firefox Build System :: General, defect, P3)

Description

[Martin Stránský [:stransky] (ni? me)]

Firefox 106.0 fails to build with GCC/PGO:

128:20.60 /builddir/build/BUILD/firefox-106.0/objdir/dist/include/nsTArray.h:656:15: error: source locations for function ‘static void AssignRangeAlgorithm<IsTriviallyCopyConstructible, IsSameType>::implementation(ElemType*, IndexType, SizeType, const Item*) [with Item = nsTString<char16_t>; ElemType = nsTString<char16_t>; IndexType = long unsigned int; SizeType = long unsigned int; bool IsTriviallyCopyConstructible = false; bool IsSameType = true]’ have changed, the profile data may be out of date [-Werror=coverage-mismatch]
128:20.60 656 | static void implementation(ElemType* aElements, IndexType aStart,

Looks like header files in objdir/dist/include/ have changed between PGO build phases.

Comment 1

[Mike Hommey [:glandium]]

Looks like header files in objdir/dist/include/ have changed between PGO build phases.

Can you give more details?

Comment 2

[Martin Stránský [:stransky] (ni? me)]

Sure. I'm testing 106.0 clang PGO build right now to see if there's any difference.

Comment 3

[Thaodan]

(In reply to Martin Stránský [:stransky] (ni? me) from comment #2)

Sure. I'm testing 106.0 clang PGO build right now to see if there's any difference.

Any results form that test?

Comment 4

[Martin Stránský [:stransky] (ni? me)]

It's buried deep on my TODO list. Will look at it when I have a chance.

Comment 5

[Thaodan]

Bug 1029346 made so that these warnings are ignored for PGO builds I wonder if that should be done now too?

Comment 6

[Martin Stránský [:stransky] (ni? me)]

(In reply to Thaodan from comment #5)

Bug 1029346 made so that these warnings are ignored for PGO builds I wonder if that should be done now too?

I made the warnings ignored but it resulted to slow PGO code so the warnings are actually correct. Still on my TODO list for further investigation.

Comment 7

[Martin Stránský [:stransky] (ni? me)]

It doesn't seem to affect performance.