RE: ld auto-import broken? (was: Re: ld auto-import)

["Ralf Habacker" <Ralf dot Habacker at freenet dot de>]

Ralf Habacker


> -----Original Message-----
> From: kde-cygwin-admin@mail.kde.org
> [mailto:kde-cygwin-admin@mail.kde.org]On Behalf Of Gerrit P. Haase
> Sent: Monday, February 04, 2002 4:24 PM
> To: Christopher January
> Cc: kde-cygwin@mail.kde.org; binutils@sources.redhat.com
> Subject: ld auto-import broken? (was: Re: ld auto-import)
>
>
> Hallo Christopher,
>
> 04.02.2002, 12:57, es wurde gemailt:
>
> >>>>> Seems to me the auto-import feature of ld is extremely broken; a separate
> >>>>> DLL reference seems to be made for every auto-imported symbol.
>
> >>>> BTW, which version of ld are you using?
>
> >>> I am using GNU ld version 2.11.90. I think it's the version from the
> >>> kde-cygwin sourceforge page.
> >>> The attached output was generated using dumpbin.exe from MSVC++.
>
> >> Are there special patches in this version?  I always use the standard
> >> cygwin version of the binutils.

Yes, it exclude all libraries from automatic exports to avoid problems with multiple linking static libs. This is
needed for kde, as otherwise kde linking fails.
Second it contains a symbol cache, for which I haven't found anyone who could check this code on the binutils list.
The symbols cache speeds up linking time and reduces needed memory while linking big libraries.  I haven't got any
problem with this cache, so it might be integrated into the official binutils, but before I would like to get
somebody who could check this patch.

>
> > See these links for the patches and README file for this version:
> > http://prdownloads.sourceforge.net/kde-cygwin/binutils_20011126.dif
> > http://prdownloads.sourceforge.net/kde-cygwin/README.binutils-20011126
>
> Hmm, then this is not the current version of the binutils ld.
> Maybe there are some changes in the current (CVS) binutils (ld)?

> At least the auto-import feature is included there for a while now.
>
> Please can you try the ld from the current Cygwin release so we'll see
> if it is the same 'problem' there.
> And can someone verify if it is a problem at all?

Ralf