minutes of ESC call ...

[Bjoern Michaelsen]

Hi,

On Thu, Aug 04, 2016 at 05:13:26PM +0100, Michael Meeks wrote:
> * writer regressions (Bjoern)
>    + since 5.0 no fully triaged, bibisected regression in writer
        https://twitter.com/Sweet5hark/status/760830653571170316
>    + if you remove filters & perf. issues down to 25 bibisected issues (of 125).
        whops, that was when I was looking at Linux only:
        https://bugs.documentfoundation.org/buglist.cgi?bug_status=NEW&component=Writer&f1=keywords&f2=short_desc&keywords=bibisected%2C%20regression%2C%20&keywords_type=allwords&list_id=625809&o1=nowordssubstr&o2=nowords&op_sys=All&op_sys=Linux%20%28All%29&order=bug_id&product=LibreOffice&query_format=advanced&resolution=---&v1=filt%20perf&v2=FILEOPEN%20FILESAVE
        without restricting to Linux its 33 bugs:
        https://bugs.documentfoundation.org/buglist.cgi?bug_status=NEW&component=Writer&f1=keywords&f2=short_desc&keywords=bibisected%2C%20regression%2C%20&keywords_type=allwords&list_id=625809&o1=nowordssubstr&o2=nowords&order=bug_id&product=LibreOffice&query_format=advanced&resolution=---&v1=filt%20perf&v2=FILEOPEN%20FILESAVE
>    + are perf. regression a big part of our load ? (Michael)
        they are not: not filtering out perf moves the count from 33 to 38
        https://bugs.documentfoundation.org/buglist.cgi?bug_status=NEW&component=Writer&f1=keywords&f2=short_desc&keywords=bibisected%2C%20regression%2C%20&keywords_type=allwords&list_id=625810&o1=nowordssubstr&o2=nowords&product=LibreOffice&query_format=advanced&resolution=---&v1=filt&v2=FILEOPEN%20FILESAVE
>        + not sure if they are, but hard to bibisect.
>            + particularly with dbgutil build bibisection ! (Miklos)
>               + do we have repositories ?
>                  + we have Windows + Mac, but not Linux. https://gerrit.libreoffice.org/#/admin/projects/?filter=bisect
>        + filter issues are the main part
         right:
         https://bugs.documentfoundation.org/buglist.cgi?bug_status=NEW&component=Writer&f1=keywords&f2=short_desc&keywords=bibisected%2C%20regression%2C%20&keywords_type=allwords&list_id=625811&o1=nowordssubstr&o2=nowords&product=LibreOffice&query_format=advanced&resolution=---
         so 92-33=59 (64%) of the fully triaged regressions are filter stuff

On an additional note: Indeed our bibisect quote could be better: e.g. in Writer 92/211 = 43%
isnt too good. 2/28 = 7% out of Writer 5.x regressions is even worse. Clearly
some room for improvement there.

Best,

Bjoern