opt_duplicate_compiler_generated_deps

Hello libtool list,

We recently came across the following bug on Fedora:
https://bugzilla.redhat.com/show_bug.cgi?id=2043517

While investigating it, we came to the conclusion that libtool seems to
wrongly deduplicate the following postdeps:

postdeps="-lstdc++ -lm -lgcc_s -lc -lgcc_s"

That leads us to the following change in principle:

diff --git a/build-aux/ltmain.in b/build-aux/ltmain.in
index e4efc925..5e8014e8 100644
--- a/build-aux/ltmain.in
+++ b/build-aux/ltmain.in
@@ -521,7 +521,7 @@ libtool_validate_options ()
     case $host in
       # Solaris2 added to fix
http://debbugs.gnu.org/cgi/bugreport.cgi?bug=16452
       # see also: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59788
-      *cygwin* | *mingw* | *pw32* | *cegcc* | *solaris2* | *os2*)
+      *cygwin* | *mingw* | *pw32* | *cegcc* | *solaris2* | *os2* | *linux*)
         # don't eliminate duplications in $postdeps and $predeps
         opt_duplicate_compiler_generated_deps=:
         ;;

In other words, and if we are not wrong, the more the time passes, the more
this case grows.
Then the question is, wouldn't it make sense to actually have
"opt_duplicate_compiler_generated_deps=:" as the default behavior, and be
done with it ?

Frédéric Bérat

Senior Software Engineer, Platform Tools

Red Hat <https://www.redhat.com>

[email protected]
<https://www.redhat.com>