Bug 39844 – sys-apps/quota fails to build on amd64

• Status: RESOLVED FIXED
• Product: Gentoo Linux

[Description Ryan C. Gordon 2004-01-29 23:03:43 UTC]

sys-apps/quota fails to build on amd64 due to a missing kernel #define.

This might be a kernel 2.6-specific issue.

Reproducible: Always
Steps to Reproduce:
1. "emerge quota"
2.
3.
Actual Results:  
Fails to compile because it can't find a #define for FIOQSIZE.

I was lame and worked around this with:

CFLAGS="$CFLAGS -DFIOQSIZE=0x5460" emerge quota

Obviously, that's not the right thing to do.




Portage 2.0.49-r21 (default-amd64-1.4, gcc-3.3.2, glibc-2.3.2-r9, 2.6.1-gentoo)
============================================================
=====
System uname: 2.6.1-gentoo x86_64 5
Gentoo Base System version 1.4.3.12
ccache version 2.3 [enabled]
ACCEPT_KEYWORDS="amd64"
AUTOCLEAN="yes"
CFLAGS="-O3 -fomit-frame-pointer -funroll-loops -pipe"
CHOST="x86_64-pc-linux-gnu"
COMPILER="gcc3"
CONFIG_PROTECT="/etc /usr/X11R6/lib/X11/xkb /usr/kde/2/share/config /usr/kde/3/share/
config /usr/share/config /var/qmail/alias /var/qmail/control"
CONFIG_PROTECT_MASK="/etc/gconf /etc/env.d"
CXXFLAGS="-O3 -fomit-frame-pointer -funroll-loops -pipe"
DISTDIR="/usr/portage/distfiles"
FEATURES="autoaddcvs ccache sandbox"
GENTOO_MIRRORS="http://gentoo.mirrors.pair.com/"
MAKEOPTS="-j3"
PKGDIR="/usr/portage/packages"
PORTAGE_TMPDIR="/var/tmp"
PORTDIR="/usr/portage"
PORTDIR_OVERLAY=""
SYNC="rsync://rsync.gentoo.org/gentoo-portage"
USE="3dnow amd64 apache apm avi berkdb crypt curl encode foomaticdb gdbm gif gpm imap 
imlib jpeg libg++ libwww maildir mbox mikmod mmx motif mpeg multilib mysql ncurses nls 
oggvorbis opengl oss pam pdflib perl php png python quicktime readline ruby slang spell sse ssl 
tcltk tcpd truetype xml xml2 xv zlib"

[Comment 1 Jason Huebel (RETIRED) 2004-02-27 21:49:09 UTC]

What version?

[Comment 2 Ryan C. Gordon 2004-02-27 23:37:07 UTC]

Not sure what version it was when I reported this, but the latest version unmasked on amd64 (sys-apps/quota-3.06-r1) still does this. Actually, the latest bleeding edge ebuild (3.10) does this, too.

I'm on the default sys-kernel/linux-headers-2.4.21, despite running a 2.6 kernel, if that's the culprit.

--ryan.

[Comment 3 Jason Huebel (RETIRED) 2004-03-30 08:01:51 UTC]

Added URL which includes (very easy) fix for FIOQSIZE issue...

[Comment 4 Jason Huebel (RETIRED) 2004-03-30 08:15:19 UTC]

Created attachment 28374 [details, diff]
Adds FIOQSIZE define required for quota to build on amd64

Patch for /usr/include/asm/ioctls.h that adds FIOQSIZE define needed by quota
on amd64.

[Comment 5 Jason Huebel (RETIRED) 2004-03-30 09:14:52 UTC]

Created attachment 28378 [details, diff]
include/asm-x86_64/ioctls.h patch

Patches include/asm-x86_64/ioctls.h to add FIOQSIZE. Needed by quote package on
amd64.

[Comment 6 Jean-Claude Repetto 2004-04-19 02:55:05 UTC]

The 3.11 version of quota-tools seems to solve the bug :

Changes in quota-tools form 3.10 to 3.11
* add FIOQSIZE for x84_64 architecture (Jan Kara)

But I could not find an ebuild for this new version.

[Comment 7 Travis Tilley (RETIRED) 2004-04-19 08:10:31 UTC]

i made an ebuild for the new version and it fails to compile.

gcc -DBSD_BEHAVIOUR -DRPC -DALT_FORMAT -DHAVE_EXT2_INO_T -O2 -ftracer -pipe -DEXT2_DIRECT -D_GNU_SOURCE -Wall -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -DQUOTA_VERSION=\"3.11\" -DHOSTS_ACCESS   -c -o quotacheck.o quotacheck.c
gcc -DBSD_BEHAVIOUR -DRPC -DALT_FORMAT -DHAVE_EXT2_INO_T -O2 -ftracer -pipe -DEXT2_DIRECT -D_GNU_SOURCE -Wall -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -DQUOTA_VERSION=\"3.11\" -DHOSTS_ACCESS   -c -o quotacheck_v1.o quotacheck_v1.c
In file included from quotaio_xfs.h:35,
                 from dqblk_xfs.h:8,
                 from quotaio.h:19,
                 from quotacheck.c:36:
/usr/include/asm/types.h:20: error: conflicting types for '__s64'
/usr/include/ext2fs/ext2_types.h:18: error: previous declaration of '__s64' was here
/usr/include/asm/types.h:21: error: conflicting types for '__u64'
/usr/include/ext2fs/ext2_types.h:19: error: previous declaration of '__u64' was here
make: *** [quotacheck.o] Error 1
make: *** Waiting for unfinished jobs....
 
!!! ERROR: sys-apps/quota-3.11 failed.

[Comment 8 Danny van Dyk (RETIRED) 2004-05-22 07:39:46 UTC]

Created attachment 31839 [details, diff]
Patch to allow sys-apps/quota to compile against linux-headers-2.4.21-r1

As the other attachments aren't available anymore (Server error 500 for me), i
readded the patch and CC'ed x86-kernel herd.

Could you kernel guys please add this little patch to linux-headers-2.4.21-r1 ?

TIA !

[Comment 9 Danny van Dyk (RETIRED) 2004-05-22 17:01:33 UTC]

Oops, dammit CC list.

x86-kernel: Would you please have a look @ above patch ?

[Comment 10 Jason Huebel (RETIRED) 2004-05-26 12:27:51 UTC]

kernel herd, please take a look at the attached patch and add it to linux-headers-2.4.21-r1. Thanks!

[Comment 11 Travis Tilley (RETIRED) 2004-07-13 18:24:49 UTC]

does 3.06-r2 still fail against 2.4 headers? it works here with 2.6.

[Comment 12 Ryan C. Gordon 2004-07-13 20:16:08 UTC]

Definitely still not working here:

gcc -DBSD_BEHAVIOUR -DRPC_SETQUOTA -DRPC -DALT_FORMAT -DHAVE_EXT2_INO_T -O3 -fomit-frame-pointer -pipe -DEXT2_DIRECT -D_GNU_SOURCE -Wall -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -DQUOTA_VERSION=\"3.06\" -DHOSTS_ACCESS   -c -o quotaops.o quotaops.c
quotacheck.c: In function `getqsize':
quotacheck.c:256: error: `FIOQSIZE' undeclared (first use in this function)
quotacheck.c:256: error: (Each undeclared identifier is reported only once
quotacheck.c:256: error: for each function it appears in.)
make: *** [quotacheck.o] Error 1



Just as a sanity check, I re-emerged sys-kernel/linux-headers-2.4.21-r1, still no luck: same error.


I noticed that there's a new glibc marked stable on amd64:

  [ebuild     U ] sys-libs/glibc-2.3.4.20040605 [2.3.2-r9]


Possibly this is why it works for you and not me?

--ryan.

[Comment 13 Jason Huebel (RETIRED) 2004-08-11 13:39:09 UTC]

Ryan, have you moved to linux26-headers yet?

[Comment 14 Tim Yamin (RETIRED) 2004-08-25 17:56:45 UTC]

I'm closing this as FIXED since AMD64 has moved over to linux26-headers, which does not have this issue. If you still experience it after switching to 2.6 headers then please reopen this bug. Thanks!