Bug 12160

Summary:	timezone %pretrans error script
Product:	Mageia	Reporter:	Manuel Hiebel <manuel.mageia>
Component:	RPM Packages	Assignee:	David Walser <luigiwalser>
Status:	RESOLVED FIXED	QA Contact:
Severity:	normal
Priority:	release_blocker	CC:	dvgevers, thierry.vignaud
Version:	Cauldron	Keywords:	NEEDINFO
Target Milestone:	---
Hardware:	i586
OS:	Linux
Whiteboard:	4RC
Source RPM:	timezone-6:2013i	CVE:
Status comment:
Bug Depends on:
Bug Blocks:	11704
Attachments:	capture of tty2 report.bug

Description Manuel Hiebel 2013-12-30 23:26:13 CET

Description of problem:

As this is shown as a popup in the installer => blocker

Reproducible: 

Steps to Reproduce:

Manuel Hiebel 2013-12-30 23:26:27 CET

Priority: Normal => release_blocker
Blocks: (none) => 11704

Comment 1 Manuel Hiebel 2013-12-30 23:27:41 CET

Created attachment 4690 [details]
capture of tty2

as install is still running, logs come latter if there is something interesting

Thierry Vignaud 2013-12-31 12:36:11 CET

CC: (none) => thierry.vignaud
Assignee: bugsquad => luigiwalser

Comment 2 Manuel Hiebel 2013-12-31 14:25:10 CET

*** Bug 12164 has been marked as a duplicate of this bug. ***

CC: (none) => dvgevers

Comment 3 Dick Gevers 2013-12-31 14:26:37 CET

:)

Log has:

* trans: scheduling update of timezone-2013i-1.mga4.i586 (id=1739, file=/tmp/image/media/core/timezone-2013i-1.mga4.i586.rpm)
* trans: scheduling update of libdrm_nouveau2-2.4.50-1.mga4.i586 (id=1560, file=/tmp/image/media/core/libdrm_nouveau2-2.4.50-1.mga4.i586.rpm)
* urpmi error: ERROR: 'script' failed for timezone-2013i-1.mga4.i586: 
* timezone not installed, %pretrans(timezone-6:2013i-1.mga4.i586) scriptlet failed, exit status 127

Comment 4 Manuel Hiebel 2013-12-31 14:27:30 CET

(In reply to Manuel Hiebel from comment #2)
> *** Bug 12164 has been marked as a duplicate of this bug. ***

I indeed forget to said for the step: 

Enable additional online media when iso has been seen: a limited number of packages starts to get installed.

certainly a missing require(pre)

Dick Gevers 2013-12-31 14:38:36 CET

URL: (none) => 4RC

Dick Gevers 2013-12-31 14:39:02 CET

URL: 4RC => (none)
Whiteboard: (none) => 4RC

Comment 5 David Walser 2014-01-01 00:23:44 CET

Nothing in this bug says what exactly's going wrong.  There's no reason that the %pretrans script in timezone should fail.  It's one command, protected by a test command, that should never fail anyway.

Comment 6 Thierry Vignaud 2014-01-01 13:53:45 CET

rm can fail.
And it has done as shown above.
Several time.

It can also be missing.
There's nothing preventing timezone to be installed before coreutils.
You're lacking a "Requires(post): coreutils"

Can reporters attach their whole /root/drakx/report.bug.xz in order to confirm?

Keywords: (none) => NEEDINFO

Comment 7 David Walser 2014-01-01 16:21:42 CET

rm -f should not fail, ever.  If it's missing, that would make sense.  rm is used in %pretrans, so it wouldn't be Requires(post).  Is there a such thing as Requires(pretrans) ?

Comment 8 Manuel Hiebel 2014-01-01 17:15:16 CET

Created attachment 4703 [details]
report.bug

Looks so for the coreutils not installed before.

Comment 9 David Walser 2014-01-09 21:13:12 CET

Should be fixed in timezone-2013i-2.mga4.

Now it tests for the existence of the directory before it tries to delete it, so if it's a fresh installation, it shouldn't exist, therefore not attempting to execute the rm command.  Previously it only tested that it wasn't a symlink, but that test returned true if it didn't exist at all.  Also fixed in Mageia 3 SVN.

Status: NEW => RESOLVED
Resolution: (none) => FIXED

Comment 10 Thierry Vignaud 2014-01-09 22:21:29 CET

You still lack a require...

Status: RESOLVED => UNCONFIRMED
Resolution: FIXED => (none)
Ever confirmed: 1 => 0

Comment 11 David Walser 2014-01-09 22:23:41 CET

(In reply to Thierry Vignaud from comment #10)
> You still lack a require...

On what?  coreutils shouldn't be needed now for new installs.

Status: UNCONFIRMED => RESOLVED
Resolution: (none) => FIXED

Comment 12 Thierry Vignaud 2014-01-09 22:54:11 CET

sight...

/me putting my rpm+urpmi maintainer hat.

That's pure non-sense.
urpmi orders packages installation (including splitting them in small transactions) according to their require/provides/conflicts tags

Beside that, their ordering is random in both:
- in the transaction (besides being ordered by their dependancies)
- between different transactions

eg, on two run of:
mkdir T2
urpmi.addmedia --urpmi-root T2 --distrib  /mageia/unstable/i586/
urpmi --urpmi-root T2 timezone --auto

one can get:
-installing perl-Carp-1.320.0-2.mga4.noarch.rpm perl-File-Sync-0.110.0-6.mga4.i586.rpm perl-base-5.18.1-3.mga4.i586.rpm libdb5.3-5.3.21-5.mga4.i586.rpm perl-Exporter-5.680.0-2.mga4.noarch.rpm libgdbm4-1.10-4.mga4.i586.rpm perl-5.18.1-3.mga4.i586.rpm libbzip2_1-1.0.6-4.mga4.i586.rpm from /mageia/unstable/i586/media/core/release
+installing libgdbm4-1.10-4.mga4.i586.rpm multiarch-utils-1.0.13-3.mga4.noarch.rpm libffi6-3.0.13-2.mga4.i586.rpm perl-base-5.18.1-3.mga4.i586.rpm libbzip2_1-1.0.6-4.mga4.i586.rpm perl-MDK-Common-1.2.29-4.mga4.noarch.rpm libpcre1-8.33-2.mga4.i586.rpm perl-File-Sync-0.110.0-6.mga4.i586.rpm from /mageia/unstable/i586/media/core/release
 Preparing...                     #############################################
-    11/32: libgdbm4              #############################################
-    12/32: libdb5.3              #############################################
-    13/32: perl                  #############################################
-    14/32: perl-base             #############################################
-    15/32: perl-Carp             #############################################
-    16/32: perl-Exporter         #############################################
-    17/32: perl-File-Sync        #############################################
-    18/32: libbzip2_1            #############################################
+    11/32: perl-base             #############################################
+    12/32: perl-File-Sync        #############################################
+    13/32: perl-MDK-Common       #############################################
+    14/32: multiarch-utils       #############################################
+    15/32: libpcre1              #############################################
+    16/32: libbzip2_1            #############################################
+    17/32: libffi6               #############################################
+    18/32: libgdbm4              #############################################

See, for the sames packages, the order differs between two runs.
The same is true, a package having very loose requires can end in any transaction after the last one that install its dependencies

So if you use use a coreutils command in some %scriptlet without any "Requires(%scriptlet): coreutils", installation will only work by chance.
But you CANNOT guaranty it'll work.

As in:

installing libpython2.7-2.7.5-11.mga4.i586.rpm libopenssl-engines1.0.0-1.0.1e-8.mga4.i586.rpm bash-completion-2.1-5.mga4.noarch.rpm timezone-2013i-1.mga4.i586.rpm python-2.7.5-11.mga4.i586.rpm libopenssl1.0.0-1.0.1e-8.mga4.i586.rpm from /mageia/unstable/i586/media/core/release
/var/tmp/rpm-tmp.EAsL5i: line 2: rm: command not found
error: %pretrans(timezone-6:2013i-1.mga4.i586) scriptlet failed, exit status 127
ERROR: 'script' failed for timezone-2013i-1.mga4.i586: 
Preparing...                     #############################################
    27/32: libopenssl1.0.0       #############################################
    28/32: libopenssl-engines1.0.0
                                 #############################################
    29/32: libpython2.7          #############################################
    30/32: python                #############################################
    31/32: bash-completion       #############################################
error: timezone-6:2013i-1.mga4.i586: install skipped

And yes this had already bitten quite a number of packagers in the years.
And yes you're just waiting for some random installation to fails on some poor users

Status: RESOLVED => REOPENED
Resolution: FIXED => (none)
Ever confirmed: 0 => 1

Comment 13 David Walser 2014-01-09 22:57:19 CET

As I already explained, the rm command from coreutils *will no longer be used* on new installations due to this change, so the Requires is not needed.

Status: REOPENED => RESOLVED
Resolution: (none) => FIXED