Bug#615157: gcc-4.6: -Wunused-but-set-variable -Werror does not imply -Werror=unused-but-set-variable
[removing the builds, not commented here]
On 28.02.2011 03:48, Jonathan Nieder wrote:
> Matthias Klose wrote:
>
>> this would require the build logs of all the *sucessful* builds from this rebuild.
>
> Indeed, "grep -F '[-Wunused-but-set'" does not dig up anything.
>
> Of 1010 failed packages:
>
> * 40 failed with
> cpp: fatal error: -fuse-linker-plugin, but liblto_plugin.so not found
>
> I think cpp is being called by rpcgen in these cases.
seen to, didn't investigate yet. but maybe recheck first with a newer gcc-4.6
build (as uploaded today).
> * 39 had missing headers, based on matching against
> /\.h: No such file or directory/. This includes things like
> /usr/include/unistd.h:227:20: fatal error: stddef.h: No such file
Most of these are ObjC builds. I assume Lucas didn't install gobjc-4.6 (and
maybe not gfortran-4.6).
> * 2 tried to use strings as char *
> (/error: invalid conversion from 'const char\*'/)
is this with -Werror? might be worth adding a FAQ how to fix these.
> * 5 had configure scripts that question the C preprocessor's sanity
> (/fails sanity check/).
I guess this is the same as the -fuse-linker-plugin error.
> * 5 suffer linking problems I didn't investigate (/undefined reference to/).
ld --no-add-needed
> * 4 c++ libraries are missing symbols (/MISSING:/).
symbols mangling
> * >= 598 seem to have had build-dependency installability problems,
> based on matching logs against /but it is not going to be installed/.
>
> Many of these are because the resolver refuses to install gcj,
> gfortran, etc from experimental unless they are mentioned as
> explicit versioned build-dependencies.
right, gcj-4.6 not available.
> * gpc4.1 redeclares YYSTYPE.
gone.
Reply to: