| Message ID | 20161206224041.GW4829@bubble.grove.modra.org |
|---|---|
| State | New |
| Headers | show |
Hi, Yeah, my SH5/SH64 removal procedures might have been a little too radical, sorry about that. However ... On Wed, 2016-12-07 at 09:10 +1030, Alan Modra wrote: > I understand that config/ in the gcc repository is the master source > for binutils-gdb config/. If that's the case then people removing > gcc support for particular targets need to be careful about their > config/ edits. If a target is still supported in binutils, then > please don't rip out config/ support for the target. How exactly is this going to work when support for a configuration has been removed from the compiler, but not from the configuration itself? That's just confusing for the user and doesn't make sense, I believe. Cheers, Oleg
On Wed, Dec 07, 2016 at 08:46:32PM +0900, Oleg Endo wrote: > Hi, > > Yeah, my SH5/SH64 removal procedures might have been a little too > radical, sorry about that. However ... > > On Wed, 2016-12-07 at 09:10 +1030, Alan Modra wrote: > > I understand that config/ in the gcc repository is the master source > > for binutils-gdb config/. If that's the case then people removing > > gcc support for particular targets need to be careful about their > > config/ edits. If a target is still supported in binutils, then > > please don't rip out config/ support for the target. > > How exactly is this going to work when support for a configuration has > been removed from the compiler, but not from the configuration itself? > That's just confusing for the user and doesn't make sense, I believe. The files in the config/ directory don't select which targets are supported by gcc. That selection is done in other files, like config.gcc. I don't see anything that would be confusing to users by leaving sh5 patterns in say config/picflag.m4. The config/* files are similar to config.guess and config.sub. Just as you do not touch config.guess and config.sub when removing a target from gcc support, so you should not need to modify config/*.
On 12/06/2016 11:40 PM, Alan Modra wrote: > I understand that config/ in the gcc repository is the master source > for binutils-gdb config/. If that's the case then people removing > gcc support for particular targets need to be careful about their > config/ edits. If a target is still supported in binutils, then > please don't rip out config/ support for the target. > > I also didn't apply one of my binutils changes to gcc's config/, > so to sync with binutils I'd like to apply the following to gcc. > gcc doesn't use AC_PROG_LEX. > > OK to apply? > > * elf.m4: Revert 2016-06-21 change. > * picflag.m4: Likewise. Revert 2016-04-30 change too. > * override.m4 (AC_PROG_LEX): Import 2016-01-18 binutils fix > for PR binutils/19481. I think this is OK for now. Have you considered removing support for those dead targets from binutils? Bernd
diff -urp ../gcc.git/config/elf.m4 config/elf.m4 --- ../gcc.git/config/elf.m4 2016-09-27 09:13:37.515759318 +0930 +++ config/elf.m4 2016-10-22 23:18:19.884410154 +1030 @@ -17,7 +17,7 @@ target_elf=no case $target in *-darwin* | *-aix* | *-cygwin* | *-mingw* | *-aout* | *-*coff* | \ *-msdosdjgpp* | *-vms* | *-wince* | *-*-pe* | \ - alpha*-dec-osf* | hppa[[12]]*-*-hpux* | \ + alpha*-dec-osf* | *-interix* | hppa[[12]]*-*-hpux* | \ nvptx-*-none) target_elf=no ;; diff -urp ../gcc.git/config/picflag.m4 config/picflag.m4 --- ../gcc.git/config/picflag.m4 2016-09-27 09:13:37.515759318 +0930 +++ config/picflag.m4 2016-10-22 23:18:19.884410154 +1030 @@ -27,6 +27,10 @@ case "${$2}" in ;; i[[34567]]86-*-mingw* | x86_64-*-mingw*) ;; + i[[34567]]86-*-interix[[3-9]]*) + # Interix 3.x gcc -fpic/-fPIC options generate broken code. + # Instead, we relocate shared libraries at runtime. + ;; i[[34567]]86-*-nto-qnx*) # QNX uses GNU C++, but need to define -shared option too, otherwise # it will coredump. @@ -57,7 +61,8 @@ case "${$2}" in $1=-fpic ;; # FIXME: Simplify to sh*-*-netbsd*? - sh-*-netbsdelf* | shl*-*-netbsdelf*) + sh-*-netbsdelf* | shl*-*-netbsdelf* | sh5-*-netbsd* | sh5l*-*-netbsd* | \ + sh64-*-netbsd* | sh64l*-*-netbsd*) $1=-fpic ;; # Default to -fPIC unless specified otherwise. diff -urp ../gcc.git/config/override.m4 config/override.m4 --- ../gcc.git/config/override.m4 2014-11-24 14:48:32.327333882 +1030 +++ config/override.m4 2016-10-22 23:18:19.884410154 +1030 @@ -101,4 +101,16 @@ m4_define([_AC_CHECK_DECLS], ]) +dnl If flex/lex are not found, the top level configure sets LEX to +dnl "/path_to/missing flex". When AC_PROG_LEX tries to find the flex +dnl output file, it calls $LEX to do so, but the current lightweight +dnl "missing" won't create a file. This results in an error. +dnl Avoid calling the bulk of AC_PROG_LEX when $LEX is "missing". +AC_DEFUN_ONCE([AC_PROG_LEX], +[AC_CHECK_PROGS(LEX, flex lex, :) +case "$LEX" in + :|*"missing "*) ;; + *) _AC_PROG_LEX_YYTEXT_DECL ;; +esac]) + ])