libtool-patches
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Libtool head: Path translation issue on MinGW


From: Ralf Wildenhues
Subject: Re: Libtool head: Path translation issue on MinGW
Date: Tue, 20 Sep 2005 14:05:27 +0200
User-agent: Mutt/1.4.1i

Hi Peter, Bob,

* Peter Ekberg wrote on Tue, Sep 20, 2005 at 10:59:23AM CEST:
> * Bob Friesenhahn wrote on Tuesday, September 20, 2005 01:52 CEST
> > On Tue, 20 Sep 2005, Peter Ekberg wrote:
> > > But that is a MinGW bug in my book. Why would ld not like a unix
> > > style path? But then again, I'm not on bleeding edge MinGW here.
> > 
> > MinGW is a native *Windows* application.  As such, it depends on
> > other applications to translate Unix style paths to Windows paths.
> > It does not know anything about Unix.  When using the MSYS shell,
> > MSYS performs that task.  MSYS does its best but apparently it does
> > not manage to detect all path arguments (difficult job) and
> > translate them.
> 
> Heh, I always thought the MinGW binaries were MSYS apps. That was a
> misconception of course. Thanks for sparing me the embarrassment of a
> "bugreport" on the MinGW list!

:-)

> > Note that the Cygwin GCC is a Cygwin application so using the Cygwin 
> > GCC in "MinGW" mode is not entirely the same.
> 
> That much was clear to me...

Side question: have you tested 'gcc -mno-cygwin' on cygwin much?
I've tried it once, but I believe I need other settings as well to
make it work smoothly.  --host?  I'm planning on eventually testing
this one thoroughly, too, though.  I know people use it.

> Anyway, here's a patch to fix this MinGW problem. Makes stresstest
> pass on MinGW if head-filter-data-symbols-3.patch is applied
> and if dumpbin is used as name lister (but the two patches are
> independent and fix separate bugs).

Looks good, please apply this one, with the really minor spelling nit
below.  Could you, by the way, backport this into branch-1-5?

>       * libltdl/m4/libtool.m4 (_LT_LINKER_SHLIBS, _LT_LANG_CXX_CONFIG):
>       Split up --out-implib option for linker in separate args so that
>       the MSYS path translation can do it's magic on the provided path

It's `its' here.  ;)

>       to the import library. Fixes problem in stresstest.at on MinGW.

Hey, we didn't invent stresstest just to please itself, these are
all bugs in libtool proper, just more or less rarely exposed.  Our
docs do not state that the output file name must not be an absolute
path, and as such, it should work to use an absolute path.  Agreed,
this one is not very likely to be hit by users.

Cheers, and thank you,
Ralf




reply via email to

[Prev in Thread] Current Thread [Next in Thread]