Attempting to build the current trunk

Tobias Burnus burnus@net-b.de
Sat Jan 4 18:43:00 GMT 2014


Damian Rouson wrote:
> Based on TomÂ’s advice (http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59654#c22) I can now build on OS X (havenÂ’t tried Linux again) after commenting out the following lines in the configure file:
>
> # Enable -Werror in bootstrap stage2 and later.
>    #enable_werror=no
> Should there be some permanent change to the configure file to avoid having to do this?

In principle, there should be no need to tweak the -Werror settings. 
When bootstrapping, the code is compiled by GCC itself (during stage 2 
and 3) and, thus, the supported diagnostics are known. If one does no 
bootstrapping, the system/specified compiler is used - which might have 
different diagnostics; thus, the compilation is done without -Werror. 
When compiling the libraries, there might be also some headers involved 
and target-dependent code - thus, e.g. libgfortran is not compiled with 
-Werror.

For the problem you are seeing (PR59663): I assume that either the patch 
was tested with --disable-bootstrap or (more likely) the 
-Werror=return-type warning was recently improved but not tested on 
Darwin before committal* - and code under gcc/config/ is target dependent.

(* That's perfectly fine: One can*not* check all target systems; the 
current development model relies on users/developers reporting bootstrap 
problems. That didn't work well as seemingly no one at macport filled a 
PR in GCC's Bugzilla.)

Regarding the issue you are seeing, I think the attached patch is the 
correct fix.

Tobias
-------------- next part --------------
A non-text attachment was scrubbed...
Name: darwin-rfc.diff
Type: text/x-patch
Size: 435 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/fortran/attachments/20140104/e0e29405/attachment.bin>


More information about the Fortran mailing list