On Sun, Jul 10, 2022 at 06:03:13PM +0100, Simon McVittie wrote: >... > On Thu, 27 Jan 2022 at 19:07:34 +0200, Adrian Bunk wrote: > > There is no defined semantics what "c-compiler" actually provides, > > e.g. 'cc' might or might not be provided. > > If a package providing c-compiler isn't guaranteed to provide any > particular interface, then what's the point of the virtual package? > I don't think we should have virtual packages that say nothing actionable > about interfaces: that's just a trap for the unwary. >... > I think a sensible place to draw the line might be that if a compiler > can compile a C program and link it to the ordinary precompiled library > stack that we ship in Debian, then it should provide both c-compiler and > the /usr/bin/cc alternative, otherwise it should provide neither of those? >...
That's still pretty vague, a reference to the specification in old POSIX/SUS versions that cc "accepts an unspecified dialect of the C language"[1] (e.g. C89) might be useful. Should "c-compiler" provide only the /usr/bin/cc alternative, or also the POSIX c89 and c99 as gcc/clang do? > smcv cu Adrian [1] https://pubs.opengroup.org/onlinepubs/7908799/xcu/cc.html
