On Wed, 2026-05-20 at 14:50 +0100, Jonathan Wakely wrote: > On Wed, 20 May 2026 at 14:09, David Malcolm <[email protected]> > wrote: > > > > Jonathan: is it OK if I push this? (and further such similar > > changes). > > It introduces an extra level of coupling in the test results > > between the > > libstdc++ headers, the optimizer, and the analyzer: changes to any > > of > > these three could lead to test failures. Arguably this is a good > > thing, > > but I wanted to give you a "heads up" about it. > > > The coupling seems fine to me, you're only relying on the public API, > and arguably if we do change some libstdc++ internals in a way that > breaks the analyzer, it would be good to know about that.
Thanks. I've pushed this as r17-627-gccfcabc847eaeb, and am continuing to poke at std::string in -fanalyzer to see what other analyzer bugs shake out. Dave
