Change-Id: Iddb724c3ae706d82933a2d82c91d07e0e36b30e3
Details
Diff Detail
- Repository
- rL LLVM
Event Timeline
docs/TableGen/LangIntro.rst | ||
---|---|---|
213–216 ↗ | (On Diff #137053) | I've found this confusing. Why would anyone expect a string argument to !isa<>() to be magically cast to a record? I would after this comment, though. :-) IMO the first line of the description alone is sufficient. |
lib/TableGen/Record.cpp | ||
1235–1238 ↗ | (On Diff #137053) | Why not use FoldingSet we use everywhere else? I can't tell whether we're in danger of pointer reuse here, though if we are we'd need both CheckType and Expr to be freed, reallocated and passed to this function simultaneously in order to trigger a problem. Still... |
lib/TableGen/TGParser.cpp | ||
958 ↗ | (On Diff #137053) | Did you mean "in !isa operator" ? |
docs/TableGen/LangIntro.rst | ||
---|---|---|
213–216 ↗ | (On Diff #137053) | Done. |
lib/TableGen/Record.cpp | ||
1235–1238 ↗ | (On Diff #137053) | Changed to use FoldingSet. Yes, the pointer reuse thing is indeed a theoretical problem, but it's pretty pervasive for all objects that can directly or indirectly reference a Record. This should really be cleaned up at some point. |
lib/TableGen/TGParser.cpp | ||
958 ↗ | (On Diff #137053) | The dangers of copy & paste :) Fixed. |