On Wed, Aug 10, 2022 at 11:30 AM Martin Liška <mli...@suse.cz> wrote: > > On 8/10/22 09:47, Richard Biener wrote: > > On Wed, Aug 10, 2022 at 9:17 AM Martin Liška <mli...@suse.cz> wrote: > >> > >> On 8/10/22 08:56, Richard Biener wrote: > >>> C++ standard library includes have to go through system.h (#define > >>> INCLUDE_STRING). > >> > >> Oh, yeah. That means I need to rely on the flat header files :/ > >> > >>> > >>> Does the API really have to use std::string? > >> > >> I would like to. My main motivation is std::string::rfind function that > >> has no C equivalent (would be rstrstr). > > > > The old code happily uses strstr though, not worrying about > > finding the last instance of --jobserver-auth? > > Yes, sorry, I forgot to mention that, it's something I was notified by the > GNU make > developer here: https://savannah.gnu.org/bugs/index.php?57242#comment13 > > > > > Anyway, I'm not going to insist - I just noticed the actual > > users use .c_str on the error message and adjusting the > > environment for a not working jobserver is done > > inconsistently. Since I'm coming from C I was more > > expecting sth like > > > > bool jobserver_active = probe_jobserver (true /* diagnose */); > > Well, the main problem is that I need to "extra" a bunch of information > when parsing the env variable (and each consumer needs something else, > so that's why the jobserver_info members). It was very ugly having all > these return values being given as params (of pointer type).
Yeah, fair enough. > Martin > > > > > rather than pulling in a class instance from an all-inline > > implementation. But hey ;) > > > > > >> > >> Martin >