Re: Bug#775155: Should provide <arch-triplet>-cc alternative
+++ Elliott Mitchell [2015-01-11 19:02 -0800]:
> Package: src:cross-gcc-defaults
> Version: 0.4
>
> The SSIA pretty much.
>
> In addition to providing /usr/bin/<arch-triplet>-gcc, the GCC
> cross-compiler packages should be providing a
> /usr/bin/<arch-triplet>-cc alternative. This is similar to a native
> compiler's /usr/bin/cc alternative, but for cross-compilers. I haven't
> yet heard plans to get CLANG/LLVM doing cross-compilation, but I imagine
> it will happen in the future and better to attack this issue before the
> pattern of using <arch-triplet>-gcc becomes entrenched and makes the
> transition harder.
You bring up an interesting point. Which is related to the issues
described on https://wiki.debian.org/ReleaseGoals/honorCCandCXX
Currently native gcc (from the gcc-defaults package) adds a 'cc'
alternative (in its postint using the standard alternatives mechanism)
(and c89-gcc c99-gcc)
Clang adds cc, c++, c89, c99 alternatives via the same route (llvm-defaults source package, clang binary package).
(notice that the c89/gcc-c89 c99/gcc-c99 alternatives do not match. Is
that a mistake?)
The obvious thing to do is do the same for the cross packages.
And for putative cross-llvm packages when someone makes some.
That would leave a gap with nothing supplying <native-triplet>-cc (to
correspond with the <native-triplet>-gcc, which gcc does already
provide) but I guess that's a simple addition to gcc-defaults.
Nothing makes alternatives change in sync, right? So one can change
'cc' to point to clang, but have 'c89' still point to gcc? And similarly
'cc' be clang but 'mips-linux-gnu-cc' be gcc? Is that OK? It's flexible.
Does anyone see any value in providing <triplet>-c89 and <triplet>-c99
links? I've never seen them used.
Copying to debian-toolchain in case anyone thinks this is not the
right thing to do?
Wookey
--
Principal hats: Linaro, Debian, Wookware, ARM
http://wookware.org/
Reply to: