This is the mail archive of the binutils@sources.redhat.com mailing list for the binutils project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: PATCH: config.if for glibc 2.2


On Wed, 16 May 2001, H . J . Lu wrote:

> Wrong. Which compiler are you using to do such a test when you build a
> cross compiler? The right compiler to use is the cross compiler you

 The one that is going to be used to build libstdc++.  At the point
config.if gets evaluated the compiler is already available and the CC
variable is set appropriately.

> are trying to configure/build. It may not even exist at this point. It
> is a chicken and egg problem. That is why I hard coded it.

 I've tested the patch building a native cross-compiler (i.e. for $build
== $host != $target) and cross-building a native compiler (i.e. for $build
!= $host == $target).  I haven't tested the cross-building a
cross-compiler case (i.e. for $build != $host != $target), but looking at
Makefiles and scripts as well as at build logs from the two former cases I
conclude CC will be set correctly as well. 

> It is a very tricky problem. There are many ugly ways to detect
> __GLIBC_MINOR__. I don't like them very much nor want to spend my
> time on them.

 Incompatible binaries are annoying enough to justify making libstdc++'s
soname consistent for both a natively built one and a cross-compiled one. 
With the current setup, depending on the configuration, you end up with
different requirements for natively built and cross-built binaries.

 I've been hit with this on MIPS/Linux; others may observe it, too. 

  Maciej

-- 
+  Maciej W. Rozycki, Technical University of Gdansk, Poland   +
+--------------------------------------------------------------+
+        e-mail: macro@ds2.pg.gda.pl, PGP key available        +


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]