Issue 105360

Summary:	corrupt rdb file on disk
Product:	udk	Reporter:	jsc
Component:	code	Assignee:	stefan.baltzer
Status:	CLOSED FIXED	QA Contact:	issues@udk <issues>
Severity:	Trivial
Priority:	P2	CC:	carsten.driesner, issues, mikhail.voytenko
Version:	DEV300m58	Keywords:	regression
Target Milestone:	OOo 3.2
Hardware:	All
OS:	All
Issue Type:	DEFECT	Latest Confirmation in:	---
Developer Difficulty:	---
Issue Depends on:
Issue Blocks:	99999, 104695

Description jsc 2009-09-25 15:19:35 UTC

extension registration process writes some registry entries in the common.rdb
(common_.rdb) in the user installation during the installation process.

- deployemnt via extension manager UI in the office  -> failed
- deployment via unopkg command line interface -> ok

Error is that the ACTIVATOR key in an implementation section is empty but should
be "com.sun.star.loader.Java2". Result is that the extension won't work later on.

After some investigation i figured out that this problem comes on the master
with m58. A sal lib from m57 under a m58 installation fixed the problem.
Probably the changes around buffered file io introduced this problem.

Comment 1 carsten.driesner 2009-09-28 09:18:50 UTC

cd: Set on CC.

Comment 2 mikhail.voytenko 2009-09-28 18:17:31 UTC

The problem is not reproducible for me in cws fwk117. All ACTIVATOR keys are
written successfully there.

mav->jsc: We will have to try the scenario on your machine to be sure.

Comment 3 mikhail.voytenko 2009-09-29 10:54:27 UTC

mav->sba: I can not reproduce the problem in fwk117. Could you please verify it.

Comment 4 mikhail.voytenko 2009-09-29 10:56:45 UTC

Just for information: the fwk117 cws contains an important fix regarding
streaming implementation in sal. Investigation of this scenario shows that this
problem could affect it.

And since the problem is not reproducible for me in fwk117 I assume that there
is a big probability that this bug is a duplicate to the one from fwk117.

Comment 5 jsc 2009-09-30 07:15:58 UTC

jsc: it seems to be a general problem that are caused by the buffered io
changes. The behaviour changed and timing problems make even more worse. I add
additional flush calls after all write operations to the registry. The registry
is the one client of these storages that are affected by the buffered io changes. 

Fixed on cws fwk121.

Comment 6 stefan.baltzer 2009-10-07 10:03:38 UTC

Verified in CWS fwk121.

Comment 7 stefan.baltzer 2009-11-05 10:21:51 UTC

OK in OOO320_m3. Closed.