Re: 3.5.3rc1 win32 / debug package ...

2012-06-13 Thread bfo

Petr Mladek wrote
 
 Create the wiki pages about getting the windows
 backtrace.
 
Hi.
Draft is available at
https://wiki.documentfoundation.org/How_to_get_a_backtrace_with_WinDbg.
LO Windows developers are asked to comment:
- what debug output do they actually need
- which additional WinDbg commands should be used (if any)
- is mini dump generationanalyse  procedure presented there useful (if not
- how to do it the right way)
Best regards.


--
View this message in context: 
http://nabble.documentfoundation.org/Re-3-5-3rc1-win32-debug-package-tp3984895p3989964.html
Sent from the Dev mailing list archive at Nabble.com.
___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-06-07 Thread Petr Mladek
Hi bfo,

I am sorry for the late replay. I had busy days.

On Mon, 2012-06-04 at 08:55 -0700, bfo wrote:
 Petr Mladek wrote
  Yes, any help is really appreciated. If you feel like, please propose
  some prioritization. 
  
 
 From my users perspective the proposal is simple:
 1. crashers
 2. dataloss
 3. regressions
 4. something do not work as expected bugs
 5. enhancements  

Nice, it looks very reasonable! I like that it is easy. I am unable to
create such easy rules myself ;-)

Well, I slightly miss some more aspects:

+ how many people are affected
+ how the problem is visible
+ workaround exists and how it is complicated
+ how old the problem is (2-3 years old bugs might loose the
  regression taste; people are getting used to the new
  behavior; we should not have such regressions but
  we sometimes end there because it is hard to debug and
  the other aspects)
  

If I compare it with the rules for Novell bugzilla, I see the following
differences:

1. Novell rules put crashers and dataloss bugs on the same level
  (critical). IMHO, it is reasonable. I guess that you would end there
   as well ;-)

2. Novell rules do not stress the word regression. They are more
   influenced by the other aspects and try to split functional problems
   into three categories (Major, Normal, Minor). I would like to have
   this in our proposal as well.

   IMHO, it does not make sense to spend time on regression that is hard
   to reproduce and fix if there are more important problems that people
   miss.

   On the other hand, it makes sense to fix behavior of a new feature
   that people like and start using heavily. It is not regression but
   it annoys quite some people.


Also we need to somehow propose rules for the priorities flag.

Would you be interested into creating more precise proposal that would
incomporate the other aspects?

We need not make too complex rules. The severity and priority are just
helper items. Developers do not strictly follow the flags. Of course,
they concentrate more on the severe bugs. Though, when they debug or
rework some piece of code, they fix even less important bugs. It is more
effective when they have the code in the head.


 BTW: With time scheduled release cycle (monthly in your case) IMHO it is
 difficult to triage bugs as blockers or critical anyway. 
 I do not know if you are prepared to do chemspills, emergency releases or
 stop the release channel if things go wrong big time...

Good point! I tried to get out of this trap by rules described at
https://wiki.documentfoundation.org/Release_Criteria#Blocker_Bug_Definition


   Maybe, but changing to NEW is quite complicated 
  in your workflow, so mostly I leave them as UNCONFIRMED.
  Heh, what is complicated about it? Is the documentation unclear? IMHO,
  if you make sure that e bug is reproducible with the given information,
  you could move it to the state NEW.
  
 
 I thought so, but according to http://wiki.documentfoundation.org/BugTriage
 there are three bold ANDs before you can change a status to NEW. 
 Further more, an exclamation mark at the end of that paragraph 
 (and the whole scary sentence before it) discouraged me to change anything
 in the Status field at all.

I wonder if the new wording is better. In each case, we need to make it
friedly. We need more people doing the bug triage.


Best Regards,
Petr

___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-06-05 Thread Bjoern Michaelsen
On Mon, Jun 04, 2012 at 07:11:03PM +0200, Nino wrote:
 I gave it a (first) try :)

Looking much better already. Thanks!

Best,

Bjoern
___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-06-04 Thread Petr Mladek
On Fri, 2012-06-01 at 13:10 -0700, bfo wrote:
 Michael Meeks-2 wrote
  
 I think it won't do any good in the long term. Enabling reporting tool would
 give you a chance to
 get more data without the need to know the bugzilla magic by the users (just
 opt in to send the reports). 
 It could bring more testers to the project, as now you have to do an own
 debug build to test LO 
 with proper backtraces (especially on Windows).
 Then (or I should write Before that) you can think about what to do with
 the data (and how). 
 Luckily you can use experience of other projects (nice reading at
 http://blog.mozilla.org/metrics/) 
 or code even and work with dashboards like https://crash-stats.mozilla.com/ 
 or http://brasstacks.mozilla.com/orangefactor for instance. 
 I can see you have quite a few  http://tinderbox.libreoffice.org tinderboxes  
 but the question is - how do you use them for QA? 
 Browsing through QA articles I can feel it is more like - download, run,
 test, report scheme atm.

I hope that we could end here. I also like Launchpad from Ubuntu. It
merges reports with the same backtrace. So, it automatically handles
duplicates. You could easily see how many people are affected and find
the related comments on a single location.

Note that we need to do this step by step. First, we need to have the
builds with debuginfo. Then we need to have infrastructure to filter the
results, handle duplicates, ... Finally, we could enable the reporting
tool.

Note that we already have troubles to filter all the manually created
bugs. We are looking for more people helping with the bug triage, see
http://wiki.documentfoundation.org/BugTriage

If you could help, please step in.

  Anyone is welcome to put / link them into whatever wiki page they
  like :-) if you want it - go for it ! I'm happy to add a link to it at
  the bottom of my template so it's easy to find in future.
  
 Yea... As wiki.d.o can have 
 http://wiki.documentfoundation.org/TDF/BoD_Meetings TDF BoD meetings  or 
 http://wiki.documentfoundation.org/TDF/Membership_Committee_Meetings
 Membership Committee Meetings  
 I simply do not understand why Engineering Steering Committee Meeting are
 not there. 
 It is very valuable information about what is going on in the project. 
 http://wiki.documentfoundation.org/RBd/TSC_Call_Minutes RBd 's notes are a
 good read, 
 but you know... pasting minutes into wiki template do not need another
 volunteer...

Heh, Michael already does so many things. In fact, he is looking for a
volunteer who could join the call and take the minutes.

Anyway, you might volunteer and take the minutes from the mailing list
and paste them into the wiki. If you prepare the pages and template and
if it is easy to add new week, Michael might do it himself in the
future.


  Also I am really concerned about your QA priorities. IMHO you should
  care little more about Windows builds and MS Office filters.
  Who is the 'you' here ? you are one of us if you're helping out :-) 
  [...]
  However, getting the best and fastest possible list of and
  prioritisation of bugs is a crucial task of QA 
  
 By you I mean QA Team.

We are still building the QA team. Feel free to join and move forward
some activities.


  Reading ESC minutes I do not see general rule of
 prioritisation of bugs.

It is pity but we do not have defined rules for prioritization of bugs.
It currently depends on the feeling of the reporters and bug triagers.

Would you be interested into creating such a proposal? You might take
inspiration in the attached document that describes rules for Novell
bugzilla. IMHO, it is reasonable and might fit even LibreOffice. We just
need to use LibreOffice-specific examples.

 Is it 1. crashers 2. regressions 3. enhancements or any other scheme? I
 can't feel it atm. 

We use the keyword regression to mark regressions. We motivate
developers to work on these bugs with higher priority.

AFAIK, we do not have any keyword for crashers. IMHO, we use a
workaround because these bugs usually have the work crash in subject.

In each case, Rainer is leading the activities in this area. I am not
sure if he has it described in the wiki. I can't find it easily.


 You have MAB meta bugs, but it's more my bug is more buggy than your bug
 attitude there.

This is only a workaround. We do not have enough people to go trough all
reported bugs and standardize the priority and severity flags. Users
tend to set higher severity for their bug, so the original setting is
not useful. MAB is the place for high priority bugs that have been
confirmed by a developer or bug triager.

MAB wont be needed once we have good rules for setting severities and
priorities and enough people who could confirm the original reports.


  Sadly I'm not QA specialist. I just decided to help LO project by
  confirming as many bugs as I can find in Bugzilla. Hope this will help
  increase the quality of LO for Windows in any way...
  Certainly - it's a really good way 

Re: 3.5.3rc1 win32 / debug package ...

2012-06-04 Thread bfo

Petr Mladek wrote
 
 If you prepare the pages and template and
 if it is easy to add new week, Michael might do it himself in the
 future.
 
Firefox team is using CreateBox Wiki Extension for minutes - check their
wiki page at https://wiki.mozilla.org/Platform#Meetings 



 Yes, any help is really appreciated. If you feel like, please propose
 some prioritization. 
 

From my users perspective the proposal is simple:
1. crashers
2. dataloss
3. regressions
4. something do not work as expected bugs
5. enhancements  

BTW: With time scheduled release cycle (monthly in your case) IMHO it is
difficult to triage bugs as blockers or critical anyway. 
I do not know if you are prepared to do chemspills, emergency releases or
stop the release channel if things go wrong big time...



  Maybe, but changing to NEW is quite complicated 
 in your workflow, so mostly I leave them as UNCONFIRMED.
 Heh, what is complicated about it? Is the documentation unclear? IMHO,
 if you make sure that e bug is reproducible with the given information,
 you could move it to the state NEW.
 

I thought so, but according to http://wiki.documentfoundation.org/BugTriage
there are three bold ANDs before you can change a status to NEW. 
Further more, an exclamation mark at the end of that paragraph 
(and the whole scary sentence before it) discouraged me to change anything
in the Status field at all.

Best regards.


--
View this message in context: 
http://nabble.documentfoundation.org/Re-3-5-3rc1-win32-debug-package-tp3984895p3988078.html
Sent from the Dev mailing list archive at Nabble.com.
___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-06-04 Thread Bjoern Michaelsen
Hi Rainer,

On Mon, Jun 04, 2012 at 08:55:59AM -0700, bfo wrote:
 I thought so, but according to http://wiki.documentfoundation.org/BugTriage
 there are three bold ANDs before you can change a status to NEW. 
 Further more, an exclamation mark at the end of that paragraph 
 (and the whole scary sentence before it) discouraged me to change anything
 in the Status field at all.

can we make that text less scary? ;)

Best,

Bjoern
___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-06-04 Thread Nino

On 04.06.2012 18:02, Bjoern Michaelsen wrote:


On Mon, Jun 04, 2012 at 08:55:59AM -0700, bfo wrote:

I thought so, but according to http://wiki.documentfoundation.org/BugTriage
there are three bold ANDs before you can change a status to NEW.
Further more, an exclamation mark at the end of that paragraph
(and the whole scary sentence before it) discouraged me to change anything
in the Status field at all.


can we make that text less scary? ;)


I gave it a (first) try :)

Nino
___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: [Libreoffice-qa] crasher bugs and most annoying bugs generated by automation [was: 3.5.3rc1 win32 / debug package ...]

2012-06-04 Thread bfo

Bjoern Michaelsen wrote
 
 Crash counter (search for libreoffice-core):
  https://errors.ubuntu.com/
 Open LibreOffice crasher bugs by most affected users:
 Open LibreOffice crasher bugs by bug heat:
 

This s very cool. Question - why I have mostly page not find errors when I
click in the Bugs report column
for libreoffice-core?

Best regards.

--
View this message in context: 
http://nabble.documentfoundation.org/Libreoffice-qa-crasher-bugs-and-most-annoying-bugs-generated-by-automation-was-3-5-3rc1-win32-debug--tp3987469p3988080.html
Sent from the QA mailing list archive at Nabble.com.
___
List Name: Libreoffice-qa mailing list
Mail address: Libreoffice-qa@lists.freedesktop.org
Change settings: http://lists.freedesktop.org/mailman/listinfo/libreoffice-qa
Problems? http://www.libreoffice.org/get-help/mailing-lists/how-to-unsubscribe/
Posting guidelines + more: http://wiki.documentfoundation.org/Netiquette
List archive: http://lists.freedesktop.org/archives/libreoffice-qa/


Re: [Libreoffice-qa] crasher bugs and most annoying bugs generated by automation [was: 3.5.3rc1 win32 / debug package ...]

2012-06-04 Thread Bjoern Michaelsen
On Mon, Jun 04, 2012 at 09:02:43AM -0700, bfo wrote:
 
 Bjoern Michaelsen wrote
  
  Crash counter (search for libreoffice-core):
   https://errors.ubuntu.com/
  Open LibreOffice crasher bugs by most affected users:
  Open LibreOffice crasher bugs by bug heat:
  
 
 This s very cool. Question - why I have mostly page not find errors when I
 click in the Bugs report column
 for libreoffice-core?

might happen if the specific bug is still private, which either is an accident
or a bug that contains confidential info (e.g. a confidential test doc).

Best,

Bjoern
___
List Name: Libreoffice-qa mailing list
Mail address: Libreoffice-qa@lists.freedesktop.org
Change settings: http://lists.freedesktop.org/mailman/listinfo/libreoffice-qa
Problems? http://www.libreoffice.org/get-help/mailing-lists/how-to-unsubscribe/
Posting guidelines + more: http://wiki.documentfoundation.org/Netiquette
List archive: http://lists.freedesktop.org/archives/libreoffice-qa/


Re: [Libreoffice-qa] crasher bugs and most annoying bugs generated by automation [was: 3.5.3rc1 win32 / debug package ...]

2012-06-02 Thread Rainer Bielefeld

Bjoern Michaelsen schrieb:


As Ubuntu, we do:
  https://errors.ubuntu.com/


boaster ;-)

Hi,

yes, I think such automatted statistics can be a good indicator.

Best regards

Rainer

___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-06-01 Thread bfo

Michael Meeks-2 wrote
 
 On Fri, 2012-05-25 at 11:35 -0700, bfo wrote:
 I read that you have unused reporting tool in the codebase and plans
 to bring it back to life. Without it do you know the top crashers?
   Not really; we rely on human bug filing and QA to bring these to our
 attention; of course, lots of users simply don't bother to file bugs,
 but hopefully we get some reports for these.
 
I think it won't do any good in the long term. Enabling reporting tool would
give you a chance to
get more data without the need to know the bugzilla magic by the users (just
opt in to send the reports). 
It could bring more testers to the project, as now you have to do an own
debug build to test LO 
with proper backtraces (especially on Windows).
Then (or I should write Before that) you can think about what to do with
the data (and how). 
Luckily you can use experience of other projects (nice reading at
http://blog.mozilla.org/metrics/) 
or code even and work with dashboards like https://crash-stats.mozilla.com/ 
or http://brasstacks.mozilla.com/orangefactor for instance. 
I can see you have quite a few  http://tinderbox.libreoffice.org tinderboxes  
but the question is - how do you use them for QA? 
Browsing through QA articles I can feel it is more like - download, run,
test, report scheme atm.



   Anyone is welcome to put / link them into whatever wiki page they
 like :-) if you want it - go for it ! I'm happy to add a link to it at
 the bottom of my template so it's easy to find in future.
 
Yea... As wiki.d.o can have 
http://wiki.documentfoundation.org/TDF/BoD_Meetings TDF BoD meetings  or 
http://wiki.documentfoundation.org/TDF/Membership_Committee_Meetings
Membership Committee Meetings  
I simply do not understand why Engineering Steering Committee Meeting are
not there. 
It is very valuable information about what is going on in the project. 
http://wiki.documentfoundation.org/RBd/TSC_Call_Minutes RBd 's notes are a
good read, 
but you know... pasting minutes into wiki template do not need another
volunteer...



 Also I am really concerned about your QA priorities. IMHO you should
 care little more about Windows builds and MS Office filters.
   Who is the 'you' here ? you are one of us if you're helping out :-) 
 [...]
   However, getting the best and fastest possible list of and
 prioritisation of bugs is a crucial task of QA 
 
By you I mean QA Team. Reading ESC minutes I do not see general rule of
prioritisation of bugs.
Is it 1. crashers 2. regressions 3. enhancements or any other scheme? I
can't feel it atm. 
You have MAB meta bugs, but it's more my bug is more buggy than your bug
attitude there.



 Sadly I'm not QA specialist. I just decided to help LO project by
 confirming as many bugs as I can find in Bugzilla. Hope this will help
 increase the quality of LO for Windows in any way...
   Certainly - it's a really good way to help out. 
 
Yes, I started doing that. It seems it moved things a little bit in few bugs
already.
Should I do something more? Maybe, but changing to NEW is quite complicated 
in your workflow, so mostly I leave them as UNCONFIRMED.
Best regards.
P.S.
Sorry for off-topic.

--
View this message in context: 
http://nabble.documentfoundation.org/Re-3-5-3rc1-win32-debug-package-tp3984895p3987465.html
Sent from the Dev mailing list archive at Nabble.com.
___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


crasher bugs and most annoying bugs generated by automation [was: 3.5.3rc1 win32 / debug package ...]

2012-06-01 Thread Bjoern Michaelsen
Hi,

On Fri, May 25, 2012 at 11:35:23AM -0700, bfo wrote:
 I read that you have unused reporting tool in the codebase and plans
 to bring it back to life. Without it do you know the top crashers? 

As Ubuntu, we do:

Crash counter (search for libreoffice-core):

 https://errors.ubuntu.com/

Open LibreOffice crasher bugs by most affected users:

 
https://bugs.launchpad.net/ubuntu/+source/libreoffice/+bugs?field.searchtext=crashorderby=-users_affected_countsearch=Searchfield.status%3Alist=NEWfield.status%3Alist=CONFIRMEDfield.status%3Alist=TRIAGEDfield.status%3Alist=INPROGRESSfield.status%3Alist=FIXCOMMITTEDassignee_option=anyfield.assignee=field.bug_reporter=field.bug_commenter=field.subscriber=field.tag=field.tags_combinator=ANYfield.status_upstream-empty-marker=1field.upstream_target=field.has_cve.used=field.omit_dupes.used=field.omit_dupes=onfield.affects_me.used=field.has_patch.used=field.has_branches.used=field.has_branches=onfield.has_no_branches.used=field.has_no_branches=onfield.has_blueprints.used=field.has_blueprints=onfield.has_no_blueprints.used=field.has_no_blueprints=on

Open LibreOffice crasher bugs by bug heat:

 
https://bugs.launchpad.net/ubuntu/+source/libreoffice/+bugs?field.searchtext=crashorderby=-heatsearch=Searchfield.status%3Alist=NEWfield.status%3Alist=CONFIRMEDfield.status%3Alist=TRIAGEDfield.status%3Alist=INPROGRESSfield.status%3Alist=FIXCOMMITTEDassignee_option=anyfield.assignee=field.bug_reporter=field.bug_commenter=field.subscriber=field.tag=field.tags_combinator=ANYfield.status_upstream-empty-marker=1field.upstream_target=field.has_cve.used=field.omit_dupes.used=field.omit_dupes=onfield.affects_me.used=field.has_patch.used=field.has_branches.used=field.has_branches=onfield.has_no_branches.used=field.has_no_branches=onfield.has_blueprints.used=field.has_blueprints=onfield.has_no_blueprints.used=field.has_no_blueprints=on

(seeing three of the top five crashers to be related to undo is a big hint in
itself -- but also no surprise to LibreOffice developers)

CC'ing the qa-list as it might be interesting for bugwranglers too -- bugs
generating heat on Ubuntu are unlikely to be harmless elsewhere.

Best,

Bjoern
___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-05-29 Thread Petr Mladek
bfo píše v Po 28. 05. 2012 v 08:37 -0700:
 bfo wrote
  
  As I wrote earlier I have pdb symbols and source links in the WinDbg
  outputs,
  but I am concerned about missing FAULTING_SOURCE_LINE and
  FAULTING_SOURCE_CODE sections.
  
 
 I managed to receive outputs with FAULTING_SOURCE_CODE section just as in
 Jesus screencast.
 FAULTING_SOURCE_LINE is still missing, but it is not a big problem, as there
 is  marker in the source code section.

Sounds great.

 I'll put it into wiki page soon.

I am looking forward to have anything there :-)

Thanks for diffing into it.


Best Regards,
Petr

___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-05-28 Thread Michael Meeks
Hi there,

On Fri, 2012-05-25 at 11:35 -0700, bfo wrote:
 I read that you have unused reporting tool in the codebase and plans
 to bring it back to life. Without it do you know the top crashers?

Not really; we rely on human bug filing and QA to bring these to our
attention; of course, lots of users simply don't bother to file bugs,
but hopefully we get some reports for these.

 Per platform? Per branch? I do not see such information in ESC minutes
  (btw: could you put them also in the wiki?).

Anyone is welcome to put / link them into whatever wiki page they
like :-) if you want it - go for it ! I'm happy to add a link to it at
the bottom of my template so it's easy to find in future.

 Also I am really concerned about your QA priorities. IMHO you should
 care little more about Windows builds and MS Office filters.

Who is the 'you' here ? you are one of us if you're helping out :-) The
priorities of participants are highly directed by what interests them,
and plenty of people are working on fixing interop. bugs / missing
interop. features / problems etc.

However, getting the best and fastest possible list of and
prioritisation of bugs is a crucial task of QA - along with finding them
though of course we have no really firm roles, many people try to do a
bit of everything. So I see UI people send patches, hackers do QA,
packagers do website work etc. etc. ;-)

 Sadly I'm not QA specialist. I just decided to help LO project by
 confirming as many bugs as I can find in Bugzilla. Hope this will help
 increase the quality of LO for Windows in any way...

Certainly - it's a really good way to help out. There is no need for
expert skills or experience, just persistence and passion :-) Thanks so
much for getting involved, having timely bug reports of windows specific
regressions, particularly those with stack traces etc. is excessively
helpful  really speeds up their fixing.

Thanks !

Michael.

-- 
michael.me...@suse.com  , Pseudo Engineer, itinerant idiot

___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-05-28 Thread bfo

bfo wrote
 
 As I wrote earlier I have pdb symbols and source links in the WinDbg
 outputs,
 but I am concerned about missing FAULTING_SOURCE_LINE and
 FAULTING_SOURCE_CODE sections.
 

I managed to receive outputs with FAULTING_SOURCE_CODE section just as in
Jesus screencast.
FAULTING_SOURCE_LINE is still missing, but it is not a big problem, as there
is  marker in the source code section.

I'll put it into wiki page soon.

Best regards.




--
View this message in context: 
http://nabble.documentfoundation.org/Re-3-5-3rc1-win32-debug-package-tp3984895p3986407.html
Sent from the Dev mailing list archive at Nabble.com.
___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-05-25 Thread Petr Mladek
Hi bfo,

it is great that you continue working on it.

Please, do not remove people from CC. There is a lot of traffic on the
mailing list. The interested people might miss your replay :-)

bfo.bugm...@spamgourmet.com píše v Čt 24. 05. 2012 v 19:56 +0200:
 I just read in ESC minutes that daily builds with --enable-symbols are
 again available at
 http://dev-builds.libreoffice.org/daily/W2008R2@20-With-Symbol-Bytemark-Hosting/master
 Unfortunately .pdb files are not available there. Maybe tinderbox 
 administrator
 could script creating an archives of pdb/source files and upload them
 just as msi installers?
 This would be great start...

Norbert, would be possible to hack the tindebox script to upload also
the .pdb file?

 I managed to get WinDbg to work with my own LOdev build with symbols
 and source on Windows 7 using VS 2008 Express.
 !analyze -v output is not exactly as on sceencast. It lacks
 FAULTING_SOURCE_LINE and FAULTING_SOURCE_CODE sections,
 but when I hit Call stack I see links to the source and by clicking it
 by hand I see used code sections as on screencast.

Sounds promising.

 Yes, I will try to draft an wiki article in few days as a starter to
 be improved by LO experts,

Great. It will help a lot if you start it.

 but I am not sure if my WinDbg output is any good.

It is probably less informative without the .pdb files. Anyway, I guess
that your are on the right way.


 In between I have to practice and browse this list more, as I still
 have problems building LO 3.5.2.2 by myself (had to disable java,
 rebasing, postgres connector). Tried with 3.5.4.2 but it is even worse
 (some cairo canvas errors).

If you are lost, just ask on this mailing list. Please, mention there
more details about your system, configure options, and the error message
with some context. Please, do it in separate mail, so we do not solve
too many problems in this thread :-)


 Maybe tinderbox operators could share their settings with the community?
 I don't think that they know all the environment configuration by
 heart. If they do, then wiki articles
 can be written in minutes :).

You might fine some information in the build_info.txt file that is
available next to the related daily build. For example, I see at
http://dev-builds.libreoffice.org/daily/W2008R2@20-With-Symbol-Bytemark-Hosting/master/2012-05-24_23.12.53/master~2012-05-24_23.12.53_build_info.txt

--- cut ---
tinderbox: administrator: nthieb...@gmail.com
tinderbox: buildname: W2008R2@20-With-Symbol-Bytemark-Hosting
tinderbox: tree: MASTER
tinderbox: pull time 2012-05-24 23:12:53
tinderbox: git sha1s
core:2733658dd47e86ac72c53565f1fd8700f884c907
dictionaries:185cbda9dd862458919ae29a70b01ae9106c85ea
help:abbbe57b0f8bd5c5b306a76d5d916da1d02cbc27

tinderbox: autogen log
running ./configure with '--with-max-jobs=12' '--with-num-cpus=8' 
'--with-vendor=Bytemark' '--disable-build-mozilla' 
'--with-ant-home=/cygdrive/c/ant/apache-ant-1.8.3' 
'--with-mozilla-build=c:/mozilla-build' '--without-junit' '--enable-symbols' 
'--disable-dependency-tracking' '--enable-ccache'
--- cut ---

Thanks for digging into it.


Best Regards,
Petr

___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-05-25 Thread Rainer Bielefeld

Petr Mladek schrieb:



Norbert, would be possible to hack the tindebox script to upload also
the .pdb file?


https://bugs.freedesktop.org/show_bug.cgi?id=50339



I managed to get WinDbg to work


Me too, with build from tinderbox 20, but gives error message missing 
symbols.


I will continue when I know how to get symbols. But I'm afraid that that 
will not work for find the reason for the lost of crashes I observe with 
latest Master Builds, performance with running WinDbg is too poor.



Best regards

Rainer
___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-05-25 Thread Petr Mladek
bfo.bugm...@spamgourmet.com píše v Čt 24. 05. 2012 v 21:20 +0200:
 On Wed, May 23, 2012 at 3:11 PM, Petr Mladek - pmla...@suse.cz wrote:
  Please consider. Always you can ask the friendly guys at Mozilla how
  to set this all up...
  Thanks for tip.
 
 What a discovery! Seems like complete how to:
 http://zenit.senecac.on.ca/wiki/index.php/Mozilla_Source_and_Symbol_Server
 Blog postings, bugs on b.m.o, other resources also worth of checking:
 http://crashopensource.wordpress.com/2007/09/20/starting-the-project-mozilla-source-and-symbol-server/
 https://bugzilla.mozilla.org/show_bug.cgi?id=385792
 https://bugzilla.mozilla.org/show_bug.cgi?id=408134
 https://bugzilla.mozilla.org/show_bug.cgi?id=419904
 https://bugzilla.mozilla.org/show_bug.cgi?id=424240
 https://bugzilla.mozilla.org/show_bug.cgi?id=424817
 https://bugzilla.mozilla.org/show_bug.cgi?id=428518
 http://web.archive.org/web/20071218162257/http://www.jorgon.freeserve.co.uk/Other/pdb.htm

Great catch. Now we need to find a volunteer who would work on this. I
have put it into bugzilla and added flags for the EasyHacks pages. 
See https://bugs.freedesktop.org/show_bug.cgi?id=50350


 It would be great if ESC could go the Mozilla's path of doing things
 in QA department.
 The knowledge base is on Mozilla's wiki. Software is... open sourced
 and available to download.
 You should get connected to make LO better. Definitely.

We are already using Litmus and are going to migrate to Moztrap. Though,
we are not familiar with the real processes that used in the Mozilla
project.

Do you have any experience with the Mozilla processes? Are you able to
compare it with the current LibreOffice processes that are described at 
http://wiki.documentfoundation.org/QA ?

This is the area where we really need improvements. It would be great if
you could join us and discus particular changes on the libreoffice-qa
mailing list.

Best Regards,
Petr

___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-05-25 Thread bfo
On Fri, May 25, 2012 at 12:03 PM, Petr Mladek - pmla...@suse.cz wrote:
 but I am not sure if my WinDbg output is any good.
 It is probably less informative without the .pdb files. Anyway, I guess
 that your are on the right way.

As I wrote earlier I have pdb symbols and source links in the WinDbg
outputs,
but I am concerned about missing FAULTING_SOURCE_LINE and
FAULTING_SOURCE_CODE sections.

 If you are lost, just ask on this mailing list. Please, mention there
 more details about your system, configure options, and the error message
 with some context. Please, do it in separate mail, so we do not solve
 too many problems in this thread :-)

Will do as I stuck on those on my way to master...
BTW: I am using sources from
http://download.documentfoundation.org/libreoffice/src/.

 You might fine some information in the build_info.txt file that is
 available next to the related daily build.

Yes, I checked that, but haven't found why I had errors in my attempts
to do the builds.

 Thanks for digging into it.

Well, it is very hard for a newcomer as more informations, settings,
parameters and tips you can get from the mailing lists than from wiki
articles. Hope to edit those a little bit when I succeed.
It is hell of a job to browse a huge number of threads to get this
running... 

Best regards.


--
View this message in context: 
http://nabble.documentfoundation.org/Re-3-5-3rc1-win32-debug-package-tp3984895p3986052.html
Sent from the Dev mailing list archive at Nabble.com.
___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-05-25 Thread bfo
On Fri, May 25, 2012 at 3:21 PM, Petr Mladek - pmla...@suse.cz wrote:
 We are already using Litmus and are going to migrate to Moztrap. Though,
 we are not familiar with the real processes that used in the Mozilla
 project.
 Do you have any experience with the Mozilla processes? Are you able to
 compare it with the current LibreOffice processes that are described at
 http://wiki.documentfoundation.org/QA ?
 This is the area where we really need improvements.

It would be difficult for me to compare it, as I am not a QA/RelEng
specialist, insider nor involved in the process.
I think you should just check what they achieved at these sites in
terms of workflow, automation and data gathering:

https://quality.mozilla.org/
https://wiki.mozilla.org/QA/Desktop_Firefox
https://wiki.mozilla.org/Socorro
https://crash-stats.mozilla.com/products/Firefox
https://tbpl.mozilla.org/

http://www.mozilla.org/hacking/module-ownership.html
https://wiki.mozilla.org/Modules
http://www.mozilla.org/hacking/committer/
http://www.mozilla.org/hacking/reviewers.html
https://wiki.mozilla.org/Sheriff_Duty
https://wiki.mozilla.org/ReleaseEngineering:Sheriffing
http://www.mozilla.org/hacking/regression-policy.html
https://wiki.mozilla.org/Inbound_Sheriff_Duty

One have to remember that Mozilla have full time QA employees and
build those tools for years.
They changed release cycle recently and also had some problems (being
so experienced).
As LibreOffice is complicated project maybe you should implement some
more automation and crash data gathering.
I read that you have unused reporting tool in the codebase and plans
to bring it back to life. Without it do you know the top crashers? 
Per platform? Per branch? I do not see such information in ESC minutes
 (btw: could you put them also in the wiki?).
Depending just on Bugzilla queries IMHO is not enough.
Also I am really concerned about your QA priorities. IMHO you should
care little more about Windows builds and MS Office filters.
Without those it will be a huge problem to gain users on this
platform. Unfortunately Windows support and MS format regressions are
little scary. I am not a developer, but Microsoft published the
documentation at
http://msdn.microsoft.com/en-us/openspecifications/cc816059.aspx. 
Unfortunately changes in the Microsoft policy about compilers
(http://arstechnica.com/information-technology/2012/05/no-cost-desktop-software-development-is-dead-on-windows-8/)
can discourage Windows volunteers or developers even more
And no - buy support, pay for code, do it yourself or change OS is not the
right answer.

 It would be great if you could join us and discus particular changes on
 the libreoffice-qa
 mailing list.

Sadly I'm not QA specialist. I just decided to help LO project by
confirming as many bugs as I can find in Bugzilla. Hope this will help
increase the quality of LO for
Windows in any way...

Best regards.


--
View this message in context: 
http://nabble.documentfoundation.org/Re-3-5-3rc1-win32-debug-package-tp3984895p3986061.html
Sent from the Dev mailing list archive at Nabble.com.
___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-05-24 Thread bfo . bugmail
On Wed, May 23, 2012 at 3:11 PM, Petr Mladek - pmla...@suse.cz wrote:
 Please consider. Always you can ask the friendly guys at Mozilla how
 to set this all up...
 Thanks for tip.

What a discovery! Seems like complete how to:
http://zenit.senecac.on.ca/wiki/index.php/Mozilla_Source_and_Symbol_Server
Blog postings, bugs on b.m.o, other resources also worth of checking:
http://crashopensource.wordpress.com/2007/09/20/starting-the-project-mozilla-source-and-symbol-server/
https://bugzilla.mozilla.org/show_bug.cgi?id=385792
https://bugzilla.mozilla.org/show_bug.cgi?id=408134
https://bugzilla.mozilla.org/show_bug.cgi?id=419904
https://bugzilla.mozilla.org/show_bug.cgi?id=424240
https://bugzilla.mozilla.org/show_bug.cgi?id=424817
https://bugzilla.mozilla.org/show_bug.cgi?id=428518
http://web.archive.org/web/20071218162257/http://www.jorgon.freeserve.co.uk/Other/pdb.htm

Best regards.

P.S.
It would be great if ESC could go the Mozilla's path of doing things
in QA department.
The knowledge base is on Mozilla's wiki. Software is... open sourced
and available to download.
You should get connected to make LO better. Definitely.

___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-05-23 Thread Petr Mladek
bfo.bugm...@spamgourmet.com píše v Pá 18. 05. 2012 v 22:22 +0200:
 On Thu, 2012-05-10 at 12:13 +0200, Jesús Corrius wrote:
  In any case, the right way to do it would be to automatically generate
  all PDBs for each released version, for example in: solver\wntmsci.pro\pdb.
  Then we put those generated PDBs online at symbols.libreoffice.org.
  I think first we should have the symbol server for Jesus' debug build.
  Or he should publish the PDB files along with the installer.
 
 Hello.
 I am Windows user interested in helping confirming UNCONFIRMED and
 other crash bug reports on b.f.o.
 Please set up the symbol server and even source server just as Mozilla did.
 This will help people without strong programming knowledge to just run
 the WinDbg,
 download the symbols/source, catch a crasher and fill complete bug reports
 for Windows platform with source excerpt included.

I remember that it has been discussed on the ESC meeting two weeks ago
or so. I do not remember the conclusion. I know that we want it. I am
not sure if we have a volunteer who could do the work. 

 And all that without setting up build environment to generate symbols
 and source access for WinDbg on their own.
 
 Links to Mozilla pages, where all this is well documented:
 https://developer.mozilla.org/en/How_to_get_a_stacktrace_with_WinDbg
 https://developer.mozilla.org/en/Using_the_Mozilla_symbol_server
 https://developer.mozilla.org/en/Using_the_Mozilla_source_server
 https://developer.mozilla.org/en/Debugging_a_minidump

Nice pointers.


 Currently there is only
 http://wiki.documentfoundation.org/Development/How_to_debug article
 and screencast by Jesús Corrius on YouTube - sadly without step by
 step wiki article about it.

The wikipage would be really helpful. Have you managed to get it working
by the screencast? Would you mind to describe the steps on the wiki? Do
not be afraid. It need not be perfect. Anyone could improve the text in
the future.

 Please consider. Always you can ask the friendly guys at Mozilla how
 to set this all up...

Thanks for tip.


Best Regards,
Petr

___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-05-20 Thread bfo . bugmail
On Thu, 2012-05-10 at 12:13 +0200, Jesús Corrius wrote:
 In any case, the right way to do it would be to automatically generate
 all PDBs for each released version, for example in: solver\wntmsci.pro\pdb.
 Then we put those generated PDBs online at symbols.libreoffice.org.
 I think first we should have the symbol server for Jesus' debug build.
 Or he should publish the PDB files along with the installer.

Hello.
I am Windows user interested in helping confirming UNCONFIRMED and
other crash bug reports on b.f.o.
Please set up the symbol server and even source server just as Mozilla did.
This will help people without strong programming knowledge to just run
the WinDbg,
download the symbols/source, catch a crasher and fill complete bug reports
for Windows platform with source excerpt included.
And all that without setting up build environment to generate symbols
and source access for WinDbg on their own.

Links to Mozilla pages, where all this is well documented:
https://developer.mozilla.org/en/How_to_get_a_stacktrace_with_WinDbg
https://developer.mozilla.org/en/Using_the_Mozilla_symbol_server
https://developer.mozilla.org/en/Using_the_Mozilla_source_server
https://developer.mozilla.org/en/Debugging_a_minidump

Currently there is only
http://wiki.documentfoundation.org/Development/How_to_debug article
and screencast by Jesús Corrius on YouTube - sadly without step by
step wiki article about it.

Please consider. Always you can ask the friendly guys at Mozilla how
to set this all up...

Best regards.

___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-05-10 Thread Jesús Corrius
On Wed, May 9, 2012 at 10:07 PM, Michael Meeks michael.me...@suse.com wrote:
 Hi there,

        Trying to get a good stack trace from vcl, I downloaded 3.5.3rc1 from:

        http://dev-builds.libreoffice.org/win32-debug/libreoffice-3-5/

        I get a beautiful crash in the MSVC++ debugger, but no symbols; I try
 to add vcllo.pdb from LOdev 3.5\program and it moans: The symbol file
 vcllo.pdb does not match the module.

In order to match the module must be from the same exact compilation.

I looked at the package and it seems *some* modules are missing
debugging symbols. Most probably something is not working right in the
build system when the debug flag is enabled.

In any case, the right way to do it would be to automatically generate
all PDBs for each released version, for example in:
solver\wntmsci.pro\pdb.
Then we put those generated PDBs online at symbols.libreoffice.org.

With that, any Microsoft debugging tool can access to the symbols with
the following path: SRV*c:\symbols*http://symbols.libreoffice.org [1]

The advantages of this approach are:

- You don't need to download a debug version.
- If you have the symbols of all released versions in the servers, the
debugging tools will find the right ones for your version
automatically.
- It's possible to create very nice stack traces automatically if
LibreOffice crashes for regular users to report.
. On the other side, if you have a server that recieves the crashes
without debugging symbols, you can make automatic prepocessing and
classification of the bugs once you get the dump from the clients in
the server side.

In any case, I will try to find some time to look at the issue in my
debug builds.

Sigh.

[1] http://support.microsoft.com/kb/311503

-- 
Jesús Corrius je...@softcatala.org
___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-05-10 Thread Michael Meeks

On Thu, 2012-05-10 at 12:13 +0200, Jesús Corrius wrote:
 In any case, the right way to do it would be to automatically generate
 all PDBs for each released version, for example in: solver\wntmsci.pro\pdb.
 Then we put those generated PDBs online at symbols.libreoffice.org.

Right ! I wonder if we can do that for master / 3.6 builds - Fridrich ?
having a symbol server:

 With that, any Microsoft debugging tool can access to the symbols with
 the following path: SRV*c:\symbols*http://symbols.libreoffice.org [1]

Would be awesome  save lots of developer time I suspect.

:-)

Michael.

-- 
michael.me...@suse.com  , Pseudo Engineer, itinerant idiot

___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-05-10 Thread Tor Lillqvist
 Then we put those generated PDBs online at symbols.libreoffice.org.

But is it documented how to set up such a symbol server? And would we
then need to bump the version numbers in the DLLs/EXEs in each and
every nightly build? Presumably the MS debuggers want to be quite sure
that some symbol file actually matches the executable being debugged.

--tml
___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-05-10 Thread Andras Timar
2012/5/10 Tor Lillqvist t...@iki.fi:
 Then we put those generated PDBs online at symbols.libreoffice.org.

 But is it documented how to set up such a symbol server? And would we
 then need to bump the version numbers in the DLLs/EXEs in each and
 every nightly build? Presumably the MS debuggers want to be quite sure
 that some symbol file actually matches the executable being debugged.

I think first we should have the symbol server for Jesus' debug build.
Or he should publish the PDB files along with the installer.

Using PDB files for release builds does not sound good to me, because
I don't think we should produce release builds with debug
configuration (i.e. no optimizations).

Andras
___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


Re: 3.5.3rc1 win32 / debug package ...

2012-05-10 Thread Michael Stahl
On 10/05/12 13:13, Andras Timar wrote:

 Using PDB files for release builds does not sound good to me, because
 I don't think we should produce release builds with debug
 configuration (i.e. no optimizations).

ah, another developer who doesn't know what our configure options do (or
are supposed to do anyway)  :)

so there are 3 options, and --enable-symbols differs from the other 2 in
that it only adds symbols to the output, it does not enable any
debugging feature, or assertion, it does not disable optimizations, and
it is primarily used not by developers, but by distributors who want to
be able to produce reliable stack traces for _product_ builds, i.e. the
Fedora RPM spec unconditionally sets --enable-symbols.

 AC_ARG_ENABLE(symbols,
 AS_HELP_STRING([--enable-symbols],
 [Include debugging symbols in output. WARNING - a complete build needs
  a lot of space (roughly 10 GiB) and takes much longer (enables -g
  compiler flag for GCC or equivalent).]),
 ,)
 
 AC_ARG_ENABLE(debug,
 AS_HELP_STRING([--enable-debug],
 [Include debugging symbols like with --enable-symbols, disable 
 compiler
  optimization and inlining plus extra debugging code like assertions.
  Extra large build! (enables -g compiler flag and dmake debug=true)
  If you need even more verbose output, build a module with
  build -- debug=true dbglevel=2.
  You can also use this switch as follows:
  --enable-debug=all -sw/ -Library_sc to enable symbols only for
  the specified gbuild-build targets (all means everything, - prepended
  means not to enable, / appended means everything in the directory,
  there is no ordering, more specific overrides more general, and
  disabling takes precedence).]))
 
 AC_ARG_ENABLE(dbgutil,
 AS_HELP_STRING([--enable-dbgutil],
 [Include additional debugging utilities, such as assertions, object
  counting, etc. Larger build. Independent from --enable-debug.
  Note that this option makes the build ABI incompatible:
  It is not possible to mix object files or libraries from a
  --enable-dbgutil and a --disable-dbgutil build.]))

___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice


3.5.3rc1 win32 / debug package ...

2012-05-09 Thread Michael Meeks
Hi there,

Trying to get a good stack trace from vcl, I downloaded 3.5.3rc1 from:

http://dev-builds.libreoffice.org/win32-debug/libreoffice-3-5/

I get a beautiful crash in the MSVC++ debugger, but no symbols; I try
to add vcllo.pdb from LOdev 3.5\program and it moans: The symbol file
vcllo.pdb does not match the module.

Thoughts ? incompetence unwinds ? it's somewhat frustrating to have a
nice, repeatable, null-pointer-dereference crasher and no way of finding
where it is ;-)

Thanks,

Michael.

-- 
michael.me...@suse.com  , Pseudo Engineer, itinerant idiot

___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice