Bug 139798 – check-reqs.eclass BUILD free space test wrong

• Status: RESOLVED FIXED
• Product: Gentoo Linux

[Description Peter Fox 2006-07-09 10:42:13 UTC]

At the beginning of the openoffice-2.0.3 build it was concerned that ${PORTAGE_TMPDIR} was less than 5120M, but the build directory is ${PORTAGE_TMPDIR}/portage which is a mount point in my system, so it is getting the test wrong.

$PORTAGE_TMPDIR is the default value (/var/tmp), ie not set in /etc/make.conf, but the build stuff actually goes into /var/tmp/portage, which has rather more room on my system.

See selected df command results:
Filesystem           1K-blocks      Used Available Use% Mounted on
/dev/hda1               972404    679708    243300  74% /
/dev/hda5              1929068    731688   1099388  40% /var
/dev/hdb1             19236308   4294184  13964972  24% /var/tmp/portage

[Comment 1 Jakub Moc (RETIRED) 2006-07-09 10:58:27 UTC]

Not an ebuild fault.

[Comment 2 Jakub Moc (RETIRED) 2006-07-10 05:15:32 UTC]

Created attachment 91351 [details, diff]
eclass/check-reqs.eclass.diff

Can you try this patch, please? :)

[Comment 3 Peter Fox 2006-07-10 09:56:50 UTC]

Doesn't work - get this:
 * Checking for at least 256MBytes RAM ...                        [ ok ]
 * Checking for at least 5120MBytes disk space at ${WORKDIR} ...  [ !! ]
 * Couldn't figure out disk space, skipping ...

[Comment 4 Jakub Moc (RETIRED) 2006-07-10 13:42:05 UTC]

Shrug... No idea why ${WORKDIR} wouldn't work.

[Comment 5 Ciaran McCreesh 2006-07-15 06:58:22 UTC]

What is the output of 'df -Pm /var/tmp/portage'?

[Comment 6 Peter Fox 2006-07-15 08:36:46 UTC]

output of 'df -Pm /var/tmp/portage':
Filesystem         1048576-blocks      Used Available Capacity Mounted on
/dev/hdb1                18786      1457     16376       9% /var/tmp/portage

[Comment 7 Ciaran McCreesh 2006-09-23 19:29:10 UTC]

${WORKDIR} doesn't exist when inside pkg_setup, so Jakub's patch is no good. There doesn't seem to be a neat solution here...

[Comment 8 Georg Sahli 2007-04-28 14:40:03 UTC]

Did run into same problem today.
Sorry for creating a dupe of this bug, had opened 176355.

[Comment 9 Andreas Proschofsky (RETIRED) 2007-04-30 12:50:54 UTC]

*** Bug 176355 has been marked as a duplicate of this bug. ***

[Comment 10 Jakub Moc (RETIRED) 2008-01-04 02:37:57 UTC]

portage folks, this is a CANTFIX until either ${WORKDIR} gets defined in pkg_setup or ${PORTAGE_TMPDIR} really means ${PORTAGE_TMPDIR} and not ${PORTAGE_TMPDIR}/portage.

[Comment 11 SpanKY 2008-01-04 05:20:56 UTC]

WORKDIR has no business existing in pkg_* functions

use T

[Comment 12 Marius Mauch (RETIRED) 2008-01-11 01:37:47 UTC]

while T probably works, the thing that's really needed here is src_check (it only applies at build time, so doesn't really belong in pkg_setup)

[Comment 13 Zac Medico 2008-01-11 03:06:09 UTC]

(In reply to comment #12)
> while T probably works, the thing that's really needed here is src_check (it
> only applies at build time, so doesn't really belong in pkg_setup)

How about src_unpack? The ebuild could do something like this:

src_unpack() {
  CHECKREQS_DISK_BUILD="2048"
  check_reqs
  unpack ${A}
}

[Comment 14 Marius Mauch (RETIRED) 2008-01-11 03:57:52 UTC]

Well, should work better than pkg_setup at least.

[Comment 15 Marius Mauch (RETIRED) 2008-03-02 15:29:54 UTC]

Not much we can do about that (adding new phase functions is beyond the scope of this)

[Comment 16 Jakub Moc (RETIRED) 2008-03-02 15:34:41 UTC]

Closing.

[Comment 17 Bo Ørsted Andresen (RETIRED) 2008-03-02 16:44:34 UTC]

Why not just use T? That's got to be better than the status quo..

[Comment 18 Bo Ørsted Andresen (RETIRED) 2008-03-03 07:53:19 UTC]

Will change it to use T.

[Comment 19 Bo Ørsted Andresen (RETIRED) 2008-04-11 13:53:59 UTC]

Committed finally. It now uses $T. Thanks everyone.