On Thu, 1 Aug 2024 at 17:57, François Dumont <frs.dum...@gmail.com> wrote: > > libstdc++: Make dg-error pattern more accurate > > Remove useless test variable and use a more accurate dg-error > pattern so > that only the ill-formed expression compilation error is considered. > > libstdc++-v3/ChangeLog: > > * testsuite/23_containers/map/operators/1_neg.cc (test01): > Remove test variable > and use 'no match' dg-error patter. > * testsuite/23_containers/set/operators/1_neg.cc (test01): > Likewise. > > Comitted with you as author.
Great, thanks. > François > > > On 01/08/2024 09:26, Jonathan Wakely wrote: > > On Thu, 1 Aug 2024 at 06:09, François Dumont <frs.dum...@gmail.com> wrote: > >> Do you want me to take care of those 2 tests ? > > Yes please. > > > >> You seem to have started something on the review of dg-error patterns in > >> use. > >> > >> Here I kept the test variable because I fear to potentially have a > >> diagnostic about unused returned value. > > I thought about that too, but you can't have an unused return value > > from a function that doesn't compile. There's no return value, because > > itr != mapByName.end() doesn't compile, so there's no function that > > gets called, so no return type. > > > Except in case of regression that would make the expression well-formed > and then produce another compilation error. But with your modification > of the dg-error pattern this cannot happen anymore so all good. >