Issue 100146 - improvement dialog does not honor -nofirststartwizard parameter, performance tests are broken
Summary: improvement dialog does not honor -nofirststartwizard parameter, performance ...
Status: CLOSED NOT_AN_OOO_ISSUE
Alias: None
Product: General
Classification: Code
Component: ui (show other issues)
Version: DEV300m42
Hardware: All All
: P2 Trivial (vote)
Target Milestone: OOo 3.2
Assignee: bjoern.michaelsen
QA Contact: issues@framework
URL:
Keywords:
Depends on:
Blocks: 90370
  Show dependency tree
 
Reported: 2009-03-12 13:17 UTC by bernd.eilers
Modified: 2009-03-20 11:24 UTC (History)
2 users (show)

See Also:
Issue Type: DEFECT
Latest Confirmation in: ---
Developer Difficulty: ---


Attachments

Note You need to log in before you can comment on or make changes to this issue.
Description bernd.eilers 2009-03-12 13:17:43 UTC
The dialog inviting the user to join the improvement program is shown even when
soffice is started with the -nofirststartwizard parameter. This currently makes
automated performance tests unusable because now soffice hangs in these tests
while showing the dialog.
Comment 1 Frank Schönheit 2009-03-13 07:52:24 UTC
I am actually not sure we should re-use the existing command line option, and
extend its semantics. Instead, we should probably a new one, since it might be
possible that both startup-dialogs need to be disabled separately.
(side note: @bei: I suppose there is a configuration option which could be
modified before starting OOo the first time, and which would disable the
improvement program invitation dialog.)
Comment 2 Olaf Felka 2009-03-13 11:13:28 UTC
of: We should think about to integrate the 'improvement dialog' to the first
start wizard.
Comment 3 Frank Schönheit 2009-03-13 11:23:11 UTC
I agree, I think this is where it actually belongs to.
Comment 4 Olaf Felka 2009-03-13 11:33:19 UTC
of: But we also need a solution for the users that doesn't start the office the
first time.
Comment 5 Frank Schönheit 2009-03-13 11:38:26 UTC
How can you use OOo if you never start it the first time? Do you start with
starting it the second time? :) Sorry, I don't get you here.
Comment 6 Olaf Felka 2009-03-13 11:51:29 UTC
The improvement dialog is a new feature in OOo 3.1. If you have installed OOo
3.0 you have started the first time wizard already. You shouldn't see him if you
start OOo 3.1.
Comment 7 Frank Schönheit 2009-03-13 12:02:43 UTC
Really? If I
- have OOo 3.0 installed
- started OOo 3.0, and finished the First Start Wizard
- install OOo 3.1
- start OOo 3.1
then I don't get the First Start Wizard? Learned something new today ... So far,
my impression was that a new installation always also triggers the First Start
Wizard, even if it works on the user data of the old installation. But I might
be wrong, admittedly ...
Comment 8 bjoern.michaelsen 2009-03-13 16:59:46 UTC
Actually, the dialog should not show up when -nofirststartwizard is used, there
is code to take care of that. Unfortuantely it seems to be broken and was a very
dirty implementation anyway as reading the cmdline params for office from an
extension (as it was back then) is ugly.

About the configuration to disable the dialog: Yes, this is possible and is
actually what is done in the smoketests. If you need a quickfix look there.

About integration in the First Start Wizard: Sounds like a good idea, however Im
not sure, if this is easily possible as the usage tracking is only available in
BUILD_SPECIAL builds and we the would need to make sure the stuff works
with/without BUILD_SPECIAL. Currently this is not a problem, as the dialog is a
job that only gets added to the configuration in BUILD_SPECIAL builds.

I will investigate that further when I return from my vacation on monday.
Comment 9 bjoern.michaelsen 2009-03-16 11:03:32 UTC
100146 blocking 90370
Comment 10 bjoern.michaelsen 2009-03-16 14:30:39 UTC
What platform and environment do you see this issue on? What is the exact
commandline used to start the Office?
I failed to reproduce this on win32 with a DEV300_m43 starting Office with:
"soffice.exe -nofirststartwizard"
As long as the -nofirststartwizard switch is set neither the wizard nor the user
feedback invitation comes up (tested multiple times, killing quickstarter
inbetween). Starting office simply with "soffice.exe" after these starts makes
the wizard and the invitation show up.
Comment 11 bernd.eilers 2009-03-16 14:52:00 UTC
bei->	b_michaelsen: I have seen this on linux DEV300_m42 during sun internal
automatic performance tests. The performance tests start the office with a
commandline like the following:
-accept=<connection_string>;urp -nocrashreport -norestore -nofirststartwizard


Comment 12 bernd.eilers 2009-03-16 17:15:01 UTC
bei->b_michaelsen: can not reproduce on DEV300_m43 on Windows either using
commandline parameters -nocrashreport -norestore -nofirststartwizard
Altough now there a different but very similar problem occurs. There´s a modal
dialog shown telling me about that I should configure the JRE under
tools/options which also now would block any automated testing.
Comment 13 bjoern.michaelsen 2009-03-19 13:12:10 UTC
@bei: Are you still being requested to set up the JRE when you remove the file
basis3.1/share/registry/modules/org/openoffice/Office/Jobs/Jobs-oooimprovement.xcu
before starting the office?
Comment 14 bjoern.michaelsen 2009-03-19 14:10:30 UTC
Is the connection string that is used to start the office on linux really
unquoted? Then its no wonder it doesnt work: The executing shell will split the
command at the semicolon and the -nofirststartwizard will never get passed to OOo.
Comment 15 bernd.eilers 2009-03-19 14:44:04 UTC
bei->b_michaelsen: The missinge quotes where just an error in the description.
In the performance test szenario a shell isn´t even involved at all so "quoting"
as a non-issue here the process is startet from a java program via Runtime.exec
which in turn I believe does just calls the C libs execl function. In the linux
process list I could verify that the hanging soffice process had indeed the
-nofirststartwizard parameter. Note that in the performance test szenario the
office is installed by calling dmake with pkg=installed parameter in
instset_native not via setup, perhaps this makes a difference.

For my other installation on windows I could not reproduce the java selection
dialog problem after I had once selected a jre and anyway I supposed that would
have to go into a separate issue as it is unrelated to the ImprovementDialog.
Comment 16 bjoern.michaelsen 2009-03-19 15:39:20 UTC
I still fail to reproduce this issue. I just installed a fresh DEV300_m43 and
started office with the following commandline again and again:
./opt/staroffice9/program/soffice "-accept=socket,host=localhost,port=8100;urp;"
-nocrashreport -norecover  -nofirststartwizard
result: no firststartwizard, no invitation to improvement program
starting office after that with:
./opt/staroffice9/program/soffice "-accept=socket,host=localhost,port=8100;urp;"
-nocrashreport -norecover 
showed the firststartwizard and the invitation dialog as expected. What is the
perfomance test doing different? I cant imagine how pkg=installed creates this
issue.

Im running out of ideas ...

As for the JRE stuff: True, thats probably an different issue.
Comment 17 bjoern.michaelsen 2009-03-19 15:51:41 UTC
Tested with linux DEV300_m42 (releng build) just to make sure. Same result as
with DEV300_m43.
Comment 18 bernd.eilers 2009-03-19 19:06:47 UTC
bei->b_michaelsen: intrestingly during the performance tests it happens quite
often that the improvement dialog is being shown but it does not happen always.
Sometimes the subtests can start the office normally without the dialog being
shown. And in no case the firststartwizard is being shown it is only the
improvement-dialog which blocks the office. In any case "ps -ef | fgrep soffice"
shows that the office was started with a  -nofirststartwizard option. We will
have to have a look into wether it might be the case that the java program
starting the performance tests errornously uses "-nofirststartwizard " (note the
trailing space) instead of "-nofirststartwizard" in some cases or something
similar. Lowering Priority to P2.




Comment 19 bernd.eilers 2009-03-20 11:22:11 UTC
bei->b_michaelsen: Further anlaysis has shown that there is indeed an additional
space in the argument. Will be closing task as invalid. Sorry for the
inconvienience.
Comment 20 bernd.eilers 2009-03-20 11:23:14 UTC
closing as invalid
Comment 21 bernd.eilers 2009-03-20 11:24:01 UTC
closing