Skip to Content.
Sympa Menu

sm-discuss - [SM-Discuss] Fw: [Openzaurus-devel] Compiler fun

sm-discuss AT lists.ibiblio.org

Subject: Public SourceMage Discussion List

List archive

Chronological Thread  
  • From: "Jose Bernardo Silva" <jbs AT bandos.homelinux.com>
  • To: "sm-discuss" <sm-discuss AT lists.ibiblio.org>
  • Subject: [SM-Discuss] Fw: [Openzaurus-devel] Compiler fun
  • Date: Tue, 11 May 2004 14:35:17 +0000

I'm not sure after reading the refered thread (and gcc.gnu.org times out for
me) that this will affect x86 architectures, but we could add the patch to our
gcc spell, no?

---------- Forwarded Message -----------
From: John Lenz <jelenz AT students.wisc.edu>
To: openzaurus-devel AT lists.sourceforge.net
Sent: Mon, 10 May 2004 21:52:12 -0500
Subject: [Openzaurus-devel] Compiler fun

I am in the progress of upgrading to 2.6.6 (well, I got the patches
upgraded). But in 2.6.6, the patches 1824 and 1825 were merged.
http://www.arm.linux.org.uk/developer/patches/viewpatch.php?id=1824/1

This is a workaround of the problem mentioned on the following thread
http://lists.arm.linux.org.uk/pipermail/linux-arm-kernel/2004-April/021526.html

Basicly, the thread says there is a problem with every gcc 3.*
including CVS. The patches check if gcc is generating invalid code and
if so, generates an error. I first encounted this bug when trying to
compile 2.6.6. The error exists for 2.6.5 (I verified it does), but
the compiler did not issue an error and stop compilation, so I never
knew about it! For some reason, it must not have caused a crash...

So beware, any 3.* version of gcc will generate wrong code when
compiling the kernel. (I have been using the cross-3.3.2 compiler from
www.openzaurus.org/official/toolchain, which definatly has this bug).

The following gcc bug report was filed for this bug
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15089
and a patch has been posted on that bug report (hasn't been merged
yet).

So I guess I have to dive in and build a toolchain. I am currently
trying to compile (using crosstool) a toolchain which includes that
patch. (It is mildly frustrating that the fix for the bug is ONE LINE,
but to recompile the compiler involves hours and lots of work :)

Have people had success compiling cross-compilers with oe?
just oebuild gcc-cross_3.3.3?
Secondly, has anyone had any success using remote Dejagnu to test the
compiler? What I am concerned about is oe has a bunch of patches to
gcc and crosstool has a bunch of different patches to gcc...

I am using crosstool because whenever anyone mentions a cross compiler
on linux-arm-kernel, everyone says to use crosstool. So I assume that
for at least kernel development, crosstool is verified to be correct.
I guess I should also try and verify that the compiler built by oe is
correct as well... Perhaps we could write an oe that uses mkjail.sh
and initjail.sh from crosstool. (Or just slightly modify crosstest.
sh)?

John

-------------------------------------------------------
This SF.Net email is sponsored by Sleepycat Software
Learn developer strategies Cisco, Motorola, Ericsson & Lucent use to deliver
higher performing products faster, at low TCO.
http://www.sleepycat.com/telcomwpreg.php?From=osdnemail3
_______________________________________________
Openzaurus-devel mailing list
Openzaurus-devel AT lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/openzaurus-devel
------- End of Forwarded Message -------


---
Jose Bernardo B R Silva | Source Mage GNU/Linux Section Guru
jbernardo AT sourcemage.org | http://www.sourcemage.org
http://counter.li.org #245602
http://cosmos.oninetspeed.pt/jose.bernardo
jbs AT bandos.homelinux.com





Archive powered by MHonArc 2.6.24.

Top of Page