[Libreoffice-qa] Need help in understanding results of bibisect

Xisco Fauli xiscofauli at libreoffice.org
Wed Oct 12 18:55:04 UTC 2016


Hi Alex,

El 12/10/16 a les 19:46, Alex McMurchy ha escrit:
> All
>
> As a training exercise I ran through a bibisect session to try and 
> track done the cause of the regression bug -
>> [Bug 102951] Mail Merge Wizard no longer works
> This was a bug that I'd raised myself, several days ago, so had some 
> interest in it.

Great! Thank you for reporting the bug and helping bibisecting it.
>
> From the bibisect I got this result -
>
>> f0be6ccdc4e4e45f3d6bcf322f98dbd66b1dcfc0 is the first bad commit
>> commit f0be6ccdc4e4e45f3d6bcf322f98dbd66b1dcfc0
>> Author: Miklos Vajna <vmiklos at collabora.co.uk>
>> Date:   Sat Mar 19 05:41:35 2016 +0100
>>
>>     2016-03-19: source-hash-b86b97e54590872fc0ea85fbea22c2d00d241181
>>
>> :100644 100644 2af32b9130734292d26f33b9ac77ed8474f735a3 
>> eb4c7262bb954e1897378ee6f73548f9dfefc234 M      build-info.txt
>> :040000 040000 bb948b1da2441f0590c9b6b90cb963ebc8f3d05c 
>> 0b17c036fcfc6e30417d0e4b52176c87fd401f1d M      opt
>
>
> I then did this
>
>>  git show f0be6cc --stat
>
> I've extracted what I believe to be relevant information, for the bug 
> report I'd raised, returned by the above command which I've recorded below
>>  opt/share/config/soffice.cfg/modules/swform/toolbar/mailmerge.xml |  
>> 25 +++++
>>  opt/share/config/soffice.cfg/modules/swreport/toolbar/mailmerge.xml 
>> |  25 +++++
>>  opt/share/config/soffice.cfg/modules/swriter/toolbar/mailmerge.xml 
>> |  25 +++++
>>  opt/share/config/soffice.cfg/modules/swriter/ui/floatingmmchild.ui 
>> |  54 -----------
>>  opt/share/config/soffice.cfg/modules/swriter/ui/mmmergepage.ui | 263 
>> --------------------------------------------------
>>  opt/share/config/soffice.cfg/modules/swriter/ui/mmoutputpage.ui | 
>> 674 
>> --------------------------------------------------------------------------------------------------------------------------------
>>  opt/share/config/soffice.cfg/modules/swriter/ui/mmpreparepage.ui | 
>> 350 ------------------------------------------------------------------
>>  opt/share/config/soffice.cfg/modules/swriter/ui/mmresultemaildialog.ui 
>> | 430 
>> ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>>  opt/share/config/soffice.cfg/modules/swriter/ui/mmresultprintdialog.ui 
>> | 292 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>>  opt/share/config/soffice.cfg/modules/swriter/ui/mmresultsavedialog.ui 
>> | 222 ++++++++++++++++++++++++++++++++++++++++++
>>  opt/share/config/soffice.cfg/modules/swxform/toolbar/mailmerge.xml 
>> |  25 +++++
>
> and
>>  160 files changed, 1126 insertions(+), 1415 deletions(-)
>
> It then dawned on me, rightly or wrongly, that /*git show f0be6cc 
> --stat*/ was showing me all the commits that had been applied on one 
> particular day.
>
> Again, rightly or wrongly, what I think I did was to go back to the 
> source code and did bisects -
>
>  1. between the commits one or two days around about Sat Mar 19
>     05:41:35 2016 the date of the first bad commit in bibisect. Found
>     nothing.
>  2. over a wider range that gave me these positive results -
>
>
>>   ba0796b 2015-12-23 | mailmerge: Killed the 'Save, print or send' page. (HEAD) [Jan Holesovsky]
>> * 6b35590 2015-12-23 | mailmerge: Implemented "Edit Individual Documents" toolbar button. (refs/bisect/good-6b355901a7e7a1fd8710e743621ad655068a5817) [Jan Holesovsky]
>> * 6d08322 2015-12-23 | mailmerge: Icon for the Mail Merge Wizard too. [Jan Holesovsky]
>> * 7342baf 2015-12-23 | mailmerge: Really show the icons for first/prev/next/last mailmerge entry. [Jan Holesovsky]
>> * d0f7cdc 2015-12-22 | mailmerge: Icons for the first/prev/next/last mailmerge entry buttons. (refs/bisect/good-d0f7cdc36d9858c3eed461377564a1875305d6e0) [Jan Holesovsky]
>> * 00fa85e 2015-12-22 | mailmerge: Make the SwMailMergeConfigItem instance long-living. [Jan Holesovsky]
>> * 906333a 2015-12-22 | mailmerge: Decrease indentation level by an early return. [Jan Holesovsky]
>> * 55202ec 2015-12-21 | mailmerge: Enable/disable the first/prev/next/last mailmerge entry buttons. [Jan Holesovsky]
>> * 827ff2e 2015-12-18 | mailmerge: Introduce buttons for first/prev/next/last mailmerge entry. [Jan Holesovsky]
>> * 299cc33 2015-12-18 | mailmerge: Don't hide the source document. [Jan Holesovsky]
>> * ac522d1 2015-12-18 | mailmerge: Introduce a mailmerge toolbar. [Jan Holesovsky]
>
> The question I've got is how one gets from the bibisect commit to the 
> actual commit in the master tree?

Mainly there are three kind of bibisect repositories:
1. 'max' repos: They have one build per commit
2. 'all' repos: They have one build made every 64 commits
3. dbgutil daily repos: They have one build made daily

So, as you're using the last one, in the end you will have a range of 
commits made between 2016-03-18 and 2016-03-19 where the problematic one 
should be.
At this point, you can run 'git bisect log', copy the output to the bug 
and add the keywords: 'regression' and 'bibisected'. More info: 
https://wiki.documentfoundation.org/QA/Bibisect/Linux#Short_.28and_sweet.29_Instructions_-_Linux_Only

However, I'd recommend you to use the max repos when possible as the 
problematic commit can be easily identified. Max repos for 5.2 are 
already available in Mac and Windows, and it should be soon for Linux too.
If you still have doubt, please ping me in IRC ( #libreoffice-qa ) and 
we can talk about it.

Regards

-- 
Xisco FaulĂ­
Libreoffice QA Team
IRC: x1sc0

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/libreoffice-qa/attachments/20161012/3200dbb4/attachment-0001.html>


More information about the Libreoffice-qa mailing list