Hi Ralf,
Den 2010-09-12 10:01 skrev Ralf Wildenhues:
>> + if { test "$len" -lt "$max_cmd_len" \
>> + || test "$max_cmd_len" -le -1; } \
>> + && ! $tool_conversion_needed_and_working
>
> `! $command' is not portable, please reverse the logic or use
> test foo = bar or so.
>
> The logic is flawed here, you are testing potentially-undefined $len.
> Did you mean to reverse the && order (the shell does short-circuiting
> of || and && so that would seem appropriate here)?
Argghh. I had one version which I scrapped pretty quickly because I saw
that problem, but I apparently forgot about it again.
>> + then
>> func_show_eval "$cmd" 'exit $?'
>
> This code has eval problems, but they were before your patch, so I guess
> they are not on you to fix.
And even if so, I would definitely not want to squash such an unrelated
and potentially problematic change in with this patch.
Anyway, here's a new attempt addressing the above if-crap and the other
nits that I didn't bother to quote.
I have tested stresstest.at on MSYS/gcc, MinGW/MSVC, Cygwin/gcc and
Cygwin/MSVC. Passes all over.
In the logs I see this on MSYS and for Cygwin/gcc:
libtool: link: dumpbin -symbols various/relative/paths ...
and
libtool: link: dumpbin -symbols /various/abs/paths ...
and on Cygwin/MSVC I see:
libtool: link: dumpbin -symbols @various/relative/paths ...
and
libtool: link: dumpbin -symbols @C:/other/abs/paths ...
So it is working as designed and not by coincidence.
Ok to push this time?
Cheers,
Peter
>From a5e4bb2eb0a47ea72ccc31a43e149a1cb6a557b8 Mon Sep 17 00:00:00 2001
From: Peter Rosin <p...@lysator.liu.se>
Date: Mon, 13 Sep 2010 09:48:57 +0200
Subject: [PATCH] Prefer $NM @file when the toolchain isn't native to $build.
* libltdl/config/ltmain.m4sh (func_mode_link): Avoid calculating
the command line length and take the @file branch *if* the file
names needs to be converted for the toolchain and the @file
branch works. Fixes problems in stresstest.at when doing a cross
from Cygwin to MinGW using MinGW native tools (a.k.a. a "faked"
cross), and for the specific case of MSVC on Cygwin it makes the
test pass.
Signed-off-by: Peter Rosin <p...@lysator.liu.se>
---
ChangeLog | 11 +++++++++++
libltdl/config/ltmain.m4sh | 21 +++++++++++++++++----
2 files changed, 28 insertions(+), 4 deletions(-)
diff --git a/ChangeLog b/ChangeLog
index c7d0336..ddd9c3e 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,14 @@
+2010-09-13 Peter Rosin <p...@lysator.liu.se>
+
+ Prefer $NM @file when the toolchain isn't native to $build.
+ * libltdl/config/ltmain.m4sh (func_mode_link): Avoid calculating
+ the command line length and take the @file branch *if* the file
+ names needs to be converted for the toolchain and the @file
+ branch works. Fixes problems in stresstest.at when doing a cross
+ from Cygwin to MinGW using MinGW native tools (a.k.a. a "faked"
+ cross), and for the specific case of MSVC on Cygwin it makes the
+ test pass.
+
2010-09-12 Peter Rosin <p...@lysator.liu.se>
* .gitignore: Ignore *.obj files for MSVC (and w32 in general).
diff --git a/libltdl/config/ltmain.m4sh b/libltdl/config/ltmain.m4sh
index 6036f4f..18f0f39 100644
--- a/libltdl/config/ltmain.m4sh
+++ b/libltdl/config/ltmain.m4sh
@@ -7349,10 +7349,23 @@ EOF
save_ifs="$IFS"; IFS='~'
for cmd1 in $cmds; do
IFS="$save_ifs"
- eval cmd=\"$cmd1\"
- func_len " $cmd"
- len=$func_len_result
- if test "$len" -lt "$max_cmd_len" || test "$max_cmd_len" -le -1;
then
+ # Take the normal branch if the nm_file_list_spec branch
+ # doesn't work or if tool conversion is not needed.
+ case $nm_file_list_spec~$to_tool_file_cmd in
+ *~func_convert_file_noop | *~func_convert_file_msys_to_w32 | ~*)
+ try_normal_branch=yes
+ eval cmd=\"$cmd1\"
+ func_len " $cmd"
+ len=$func_len_result
+ ;;
+ *)
+ try_normal_branch=no
+ ;;
+ esac
+ if test $try_normal_branch = yes \
+ && { test "$len" -lt "$max_cmd_len" \
+ || test "$max_cmd_len" -le -1; }
+ then
func_show_eval "$cmd" 'exit $?'
skipped_export=false
elif test -n "$nm_file_list_spec"; then
--
1.7.1
