In bsdgrep(1), fix escape map building in the regex parser. It was previously using memory not explicitly initialized, and the MBS escape map was being built based on a version of the pattern with escapes already parsed out.
PR: 175314
Details
Details
Run test cases stated by jbeich@ in the PR, ensure no further failures in kyua tests
Diff Detail
Diff Detail
- Repository
- rS FreeBSD src repository - subversion
- Lint
Lint Not Applicable - Unit
Tests Not Applicable
Event Timeline
Comment Actions
- Update following r316492 to remove the "xmalloc" bits
- Fix bogus indentation with ts=8
Comment Actions
- Consistently test pointers against NULL in this section
- As a minor optimization, if we hit an unescaped dot in the pattern, break out after setting firstdot if we're not building an escmap. This will also hopefully make it more clear upon later review why we bother entering this section at all if fg->hasdot but fg->wescmap == NULL
Comment Actions
- Revert bogus optimization; I missed a bit in one of the latter macros (FILL_QSBC => _FILL_QSBC, but not _FILL_QSBC_REVERSED) where hasdot is relied upon to point to the final unescaped dot
Comment Actions
It's worth noting that I don't have much faith at the moment that (_escmap/fg->wescmap/fg->escmap in !TRE_WCHAR case) construction is correct, either. I'm fairly certain it should be taking into account escape sequences, but I will check that out and address it in a separate patch.
Comment Actions
Apologies, I meant to update this sooner -- I'm kind of torn here. This may no longer be necessary if we can run D10282 through or go a step further and just rip out the TRE bits instead of the intermediate D10282. I feel fairly confident that our regex(3) implementation can much more reasonably fulfill this duty, and just as quickly.
Comment Actions
This may no longer be necessary if we can run D10282.
I'd rather we commit the work that's already been done here, even if we eventually remove this altogether.