On Tue Feb 4 17:34:19 2025 +0000, Paul Gofman wrote:
since the two MR clearly overlap (yet complement each other), how to
you want to proceed? shall we merge them into a single one ? I think that MR has a bigger part which is unrelated here, probably merging won't increase the chances for this to get in. You are already introducing a special memory allocation handling somewhat similar to the last patch in that MR. I'd suggest that maybe you can incorporate tests from there: HeapSize at correct offset (so that matches Windows and your implementation) and element count (if applicable, maybe there is no such for tls array entries you are changing)?
ok will do... (need to test for the count in the vector... I'd be tempted to say yes as it really looks like a vector layout with array of elements at ptr, and control structure just below ptr... maybe ptr[-1] is size of element - wild guess here- or unaligned ptr if needed)