Stubdom compilation failure on Fedora 24 beta

Stubdom compilation failure on Fedora 24 beta

[Marek Marczykowski-Górecki]

[-- Attachment #1.1: Type: text/plain, Size: 1734 bytes --]

Hi,

I'm trying to build Xen 4.6.1 on Fedora 24 beta. Since gcc 6, I need to pull
some patches from unstable, but then I hit some strange problem:

During configure run in stubdom/gmp-x86_64 I've got:

    checking size of unsigned short... configure: error: cannot compute sizeof (unsigned short)
    See `config.log' for more details.

It turned out the failing line in conftest is:

    return ferror (f) || fclose (f) != 0;

ferror(f) returned 1.
That's strange, when running under gdb, manually checking ferror(f) (`p
ferror(f)`) returns 0. So I've looked at code produced (after slight
modification to have "if (ferror(f)) ..."), ferror(f) was turned to:

    0x000000000040063d <+55>:    testb  $0x40,0x10(%rbx)

This doesn't look like ferror from glibc. And 0x40 is not _IO_ERR_SEEN (0x20).

This looks like ferror from newlib (stubdom/newlib-1.16.0/newlib/libc/include/stdio.h):

    #define __SERR  0x0040      /* found error */
    #define __sferror(p)    (((p)->_flags & __SERR) != 0)
    #define ferror(p)   __sferror(p)


Configure is called this way (stubdom/Makefile):
    cd $@; CPPFLAGS="-isystem $(CROSS_PREFIX)/$(GNU_TARGET_ARCH)-xen-elf/include $(TARGET_CPPFLAGS)" CFLAGS="$(TARGET_CFLAGS)" CC=$(CC) $(GMPEXT) ./configure --disable-shared --enable-static --disable-fft --without-readline --prefix=$(CROSS_PREFIX)/$(GNU_TARGET_ARCH)-xen-elf

And indeed $(CROSS_PREFIX)/$(GNU_TARGET_ARCH)-xen-elf/include contains headers
from newlib.

Any idea how to fix this? And why it was working before?

-- 
Best Regards,
Marek Marczykowski-Górecki
Invisible Things Lab
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 473 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: Stubdom compilation failure on Fedora 24 beta

[Wei Liu]

Hi Marek

On Fri, May 13, 2016 at 02:35:21PM +0200, Marek Marczykowski-Górecki wrote:
> Hi,
> 
> I'm trying to build Xen 4.6.1 on Fedora 24 beta. Since gcc 6, I need to pull
> some patches from unstable, but then I hit some strange problem:
> 
> During configure run in stubdom/gmp-x86_64 I've got:
> 
>     checking size of unsigned short... configure: error: cannot compute sizeof (unsigned short)
>     See `config.log' for more details.
> 
> It turned out the failing line in conftest is:
> 
>     return ferror (f) || fclose (f) != 0;
> 
> ferror(f) returned 1.
> That's strange, when running under gdb, manually checking ferror(f) (`p
> ferror(f)`) returns 0. So I've looked at code produced (after slight
> modification to have "if (ferror(f)) ..."), ferror(f) was turned to:
> 
>     0x000000000040063d <+55>:    testb  $0x40,0x10(%rbx)
> 
> This doesn't look like ferror from glibc. And 0x40 is not _IO_ERR_SEEN (0x20).
> 
> This looks like ferror from newlib (stubdom/newlib-1.16.0/newlib/libc/include/stdio.h):
> 
>     #define __SERR  0x0040      /* found error */
>     #define __sferror(p)    (((p)->_flags & __SERR) != 0)
>     #define ferror(p)   __sferror(p)
> 

Yes, it should be from newlib. This is expected because stubdom uses
newlib after all.

> 
> Configure is called this way (stubdom/Makefile):
>     cd $@; CPPFLAGS="-isystem $(CROSS_PREFIX)/$(GNU_TARGET_ARCH)-xen-elf/include $(TARGET_CPPFLAGS)" CFLAGS="$(TARGET_CFLAGS)" CC=$(CC) $(GMPEXT) ./configure --disable-shared --enable-static --disable-fft --without-readline --prefix=$(CROSS_PREFIX)/$(GNU_TARGET_ARCH)-xen-elf
> 
> And indeed $(CROSS_PREFIX)/$(GNU_TARGET_ARCH)-xen-elf/include contains headers
> from newlib.
> 
> Any idea how to fix this? And why it was working before?
> 

You're not the first one that got bitten by this.

Search for

Error compiling Xen in configuring stubdom/gmp with glibc-2.23 on Arch Linux

or message-id

<1457311127.3237075.541322218.485CFC99@webmail.messagingengine.com>

in xen-users@ archive.

The thread contains some analysis, but it's not yet complete.

There are two possibilities that I can think of off the top of my head:
1. the cross-build environment is buggy 2. newlib bit-rots too much.

The first requires fixing the build environment, the second requires
patching newlib (preferable imo) or upgrading it to a newer version.

It's on our radar but Ian and I are working on other things at the
moment.

Wei.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel