It's possible a note is not properly padded, which yields a short copy.
Rather than fail, just copy as much as possible, and complete it. This
works around problems found in mesa with AMD GPU pipeline.
Details
Details
- Reviewers
emaste markj jkoshy_users.sourceforge.net
Diff Detail
Diff Detail
- Repository
- rS FreeBSD src repository - subversion
- Lint
Lint Skipped - Unit
Tests Skipped - Build Status
Buildable 38652 Build 35541: arc lint + arc unit
Event Timeline
Comment Actions
Can you provide more info about the problem? Does it arise because of the roundup of namesz and descsz?
| contrib/elftoolchain/libelf/libelf_convert.m4 | ||
|---|---|---|
| 27 | Since elftoolchain aims to be portable, I suspect depending on param.h for MIN() is not really desirable. | |
| 972 | Should _libelf_cvt_NOTE_tof() be modified too? | |
Comment Actions
If I remember correctly (it's been a long time since I made the change), yes, it's the roundup that's the problem. I looked at how the GNU equivalent also only copies as much as is available. I didn't understand why it needed to be sized on a word boundary, because it's just doing a memcpy(), it's not swapping bytes in the copy. I guess because they must be properly aligned?
| contrib/elftoolchain/libelf/libelf_convert.m4 | ||
|---|---|---|
| 27 | Sure, I can make it local. | |
| 972 | Probably. I didn't bother with it because it wasn't needed for Mesa. | |