59322 - XPInstall fails on JRE: miscalculates available disk space

Status: VERIFIED DUPLICATE of bug 71393

(Core Graveyard :: Installer: XPInstall Engine, defect, P1)

Description

[Dan Rosen]

from install.log:

-------------------------------------------------------------------------------
file:///tmp/.tmp.xi.0/bin/../jre.xpi  --  11/06/2000 19:45:35
-------------------------------------------------------------------------------

     Sun Java 2
     ----------

     ** initInstall: 0
     ** plugins folder: /home/dr/ns6/plugins/
     ** Insufficient disk space: /home/dr/ns6/plugins/
     **   required : 40098 K
     **   available: 27000 K

     Install **FAILED** with error -235

I'm guessing that this is either a problem with nsIFile or with the fact that my
machine is running Redhat 7 (which has given me endless trouble...). I have 4
gigabytes of space available in my /home partition, with no quotas on that
filesystem or anything. See:

dr@midget ~ -> df -k
Filesystem           1k-blocks      Used Available Use% Mounted on
/dev/hda5              5045104    568092   4220732  12% /home

This appears to be misread by the installer as 27 megs.

Comment 1

[Dan Rosen]

per sgehani, cc'ing dveditz

Comment 2

[Samir Gehani]

Assessed as not a stop-ship bug because this showed up on a RH 7 system only so
far (although it is theoretically a bug on lower kernel revs too but there is
insufficient empirical evidence to confirm the latter).  Must be fixed by next
release.  Needs more investigation.

Comment 3

[Jimmy Lee]

Ccing David and Paul.

Comment 4

[Jimmy Lee]

Nominating.

Comment 5

[Syd Logan]

Naw, I bet the installer is looking for space in TMPDIR which is probably filled
up or too small on your system... hrmmm, this happened over a year ago (boy, we
are quick to respond, eh? :-) probably can ask for a full du -k output at this
point?

Comment 6

[Jimmy Lee]

I think we need to address this one.  I don't know in what time frame, but I
keep hearing about this type of problem every now and then.

Comment 7

[Syd Logan]

reassigning to dprice. nsIFile may be returning bogus values on large drives,
some kind of overflow problem perhaps?

Comment 8

[Syd Logan]

comment

Comment 9

[dprice (gone)]

What version of redhat are you using?  I'm testing on redhet 7.1 and I'm not
seeing it.  I wrote a simple script that calls our File.diskSpaceAvailable and
it produces the following output:

-------------------------------------------------------------------------------
file:///home/dprice/avail.xpi  --  02/20/2002 19:37:48
-------------------------------------------------------------------------------
 
     Functional: diskSpaceAvailable
     ------------------------------
 
     ** File.diskSpaceAvailable reports 8363941888
 
     Install completed successfully
     Finished Installation  02/20/2002 19:37:48
 

This is consistant with the output from df

7:37pm dp @h-10-169-103-115/export/60Trunk/mozilla/dist/bin > df -k
Filesystem           1k-blocks      Used Available Use% Mounted on
/dev/hdb1              5044156   1271392   3516532  27% /
/dev/hdb5             10080488   1400504   8167916  15% /export

Comment 10

[Daniel Veditz [:dveditz]]

This was fixed with revision 1.44 of nsLocalFileUnix.cpp on 2001/04/18, checked
in by me as it turns out (old code had dodgy long long usage).

*** This bug has been marked as a duplicate of 71393 ***

Comment 11

[Jimmy Lee]

Marking Verified since Bug 71393 is verified.