On Sun 2026-05-03 08:59:10, Jeffrey Law wrote: > > > On 4/29/2026 12:16 AM, Richard Biener wrote: > > > The primary difference is we've found #ifdef style code tends to create > > > long > > > term headaches. You end up with conditionally compiled code and it's too > > > easy > > > to make mistakes and end up with unused variables in some contexts and > > > other > > > problems. So we've generally been moving towards either runtime checks. > > > > > > I would tend say see this case as fairly minor, but it's a slippery slope > > > and > > > I'm not keen to give back the progress we've made on this front over the > > > last > > > decade or so. So if we can make it a runtime test selection rather than > > > #ifdefs, that would be preferred. > > But it's for a compile-time (of GCC) decision (instrumenting GCC itself > > with ASAN)? So I'm a bit confused. > True. I was hoping we could find a way to make this behave more like how we > handle -fchecking to make it a runtime selection rather than a compile-time > selection. If we can't, then so-be-it and Filip's patch is probably the > only sensible path forward. > > jeff
As per > >> I'm not excited by new #ifdefs. I don't guess we have any runtime way to > >> check for asan being enabled? Assuming the answer to that is no, we don't > >> have a way to check that, then this is OK for gcc-17. I've committed the patch. Hopefully that's ok. I really don't believe there is a reasonable way to make this into a run-time check. Cheers, Filip Kastl
