Review Request 110633: Don't report crashes if version is disabled in Bugzilla
Jekyll Wu
adaptee at gmail.com
Sat May 25 10:19:37 UTC 2013
> On May 24, 2013, 9:44 p.m., Jekyll Wu wrote:
> > Bugzilla itself (since 4.2.5) already rejects any attempt against disabled versions. So even without this patch, DrKonqi users won't be able to create crash report against disabled versions in the end. From developers POV, you don't need to worry about that.
> >
> > The problem is usability for users. I'm not sure this reused error dialog is more informative than the existing one in https://bugs.kde.org/attachment.cgi?id=78600&action=edit. So I'm against this patch in its current simple form.
> >
> > As said in [1][2], I'm working on a patch for the usability improvement and plan to make it into 4.11. I will create a review request today or tomorrow.
> >
> >
> > [1] https://bugs.kde.org/show_bug.cgi?id=315073#c3
> > [2] https://bugs.kde.org/show_bug.cgi?id=318769#c1
>
> Martin Gräßlin wrote:
> > Bugzilla itself (since 4.2.5) already rejects any attempt against disabled versions.
> Bugzilla does blcok, but DrKonqi still reports them as "unknown" version. I know it because I see the crash reports coming in
>
> Martin Gräßlin wrote:
> > So I'm against this patch in its current simple form.
> Are you also against into backporting it to prevent that we stop getting useless crash reports? I worked on this for fixing a real world problem. Just look at:
> https://bugs.kde.org/buglist.cgi?list_id=661927&bug_severity=crash&chfieldto=Now&query_format=advanced&chfield=[Bug%20creation]&chfieldfrom=2013-01-01&version=unspecified&longdesc=kwin%20%284.8&product=kwin&longdesc_type=allwordssubstr
>
> All crashes reported this year against KWin 4.8.
>
> Yes the dialog might not be best, but I cannot change it because it's in string freeze. I agree that for 4.11 something better should be done, but this is more for 4.10 and older (especially older).
>
> Ben Cooksley wrote:
> Perhaps it might be worth requesting a string freeze exemption here?
>
> Martin Gräßlin wrote:
> > Perhaps it might be worth requesting a string freeze exemption here?
> We would need that from each and every distribution. It doesn't help that KDE did the freeze exception and Debian is then not accepting the patch because it violates their policy.
>
> Jekyll Wu wrote:
> Those reports are all from KDE SC 4.8.5, where kwin version (using KDE_VERSION_STRING) was in the strange form of "4.8.5 (4.8.5)" . For some strange reason I'm still not aware of, bugs.kde.org fails to reject reports with that kind of version (probably due the space in the version) as it should have. I'm well aware of those reports which shouldn't have been accepted at the first place (since I'm subscribed to kde-bugs-dist@), but I haven't got the time to do a good investigation. Anyway, those "escaped" crash reports are definitely the race cases.
>
>
> I understand you are annoyed by those kwin crash reports from 4.8.5, but backporting this simple patch won't help you much: the DrKonqi from KDE SC 4.8.x simply don't contain the code of fetching version info from bugs.kde.org through the Product.get webservice API, so backing this simple patch to 4.8.x is useless. I added those code not long time ago, as the necessary base of my final goal of preventing outdated crash reports.
>
> So if your goal is not seeing those kwin 4.8.5 crash reports anymore, there are two quick solutions:
>
> 1. apply this simple drkonqi patch to 4.10.x, then ask distribution packagers(I'm think of Debian Stable) to ship DrKonqi from KDE SC 4.10.4 together with other components from KDE SC 4.8.4/5. I'm not sure how distribution packagers will think of this kind of mixture.
>
> 2. prepare a one-line patch to make kwin from KDE 4.8.5 use version "4.8.5" explicitly, instead of using KDE_VERSION_STRING. That way, bugs.kde.org itself will reject those crash reports as it has done for 4.9.{0,1,2,3} . Then ask distribution packers to apply this kwin one-line patch. I think this is more acceptable for distribution packagers. Or maybe that one line patch should be created against kdelibs, changing KDE_VERSION_STRING to the plain "4.8.5" ?
>
>
> Martin Gräßlin wrote:
> > Or maybe that one line patch should be created against kdelibs, changing KDE_VERSION_STRING to the plain "4.8.5" ?
> That's the place where it needs to be done, I think. KWin doesn't set a version number it uses KDE_VERSION_STRING. On the other hand fixing in kdelibs won't fix the problem as it would require to recompile everything.
>
> This is btw. not just a problem for KWin (that's just what I am familiar with), it applies to all software from KDE SC. The problem of 4.8.4 and 4.8.5 is common in the complete workspaces and it's not going to go away as it's the version used in Debian Stable and Ubuntu LTS.
>
> Would creating versions "4.8.4 (4.8.4)" and "4.8.5 (4.8.5)" in bugzilla fix the problem?
>
> Jekyll Wu wrote:
> I already created and disabled those (kwin) versions in bugs.kde.org some weeks ago, thinking that would work. Unfortunately, that doesn't prevent https://bugs.kde.org/show_bug.cgi?id=320106 . I'm not 100% sure who(drkonqi or bugs.kde.org) is at fault, but my guess is bugs.kde.org, since drkonqi filled all other information (product, platform, severity,etc) correctly, which means bugs.kde.org happily accepted the very first request sent by drkonqi.
>
> Martin Gräßlin wrote:
> There must be something else to solve this riddle. If it works for other versions than there must be some difference in the handling.
>
> Ben Cooksley wrote:
> I'll investigate post_bug.cgi on Bugzilla - I have seen code which causes it to fallback to "good" values when it receives a request previously I think.
>
> Ben Cooksley wrote:
> I've just checked, my guess is that Dr Konqi is guessing the list of versions from somewhere - but of course the "disabled" ones aren't shown in the output it is probably relying upon. Bugzilla definitely has numerous filters to block non-existant or disabled versions.
> Ben Cooksley 14 minutes ago (May 25, 2013, 9:32 a.m.)
> I've just checked, my guess is that Dr Konqi is guessing the list of versions from somewhere - but of course the "disabled" ones aren't shown in the > output it is probably relying upon. Bugzilla definitely has numerous filters to block non-existant or disabled versions.
Well, DrKonqi uses Product.get API to fetch product information from bugs.kde.org, including all available versions (either active or disabled). But that is not the point of those "escaped" reports which shouldn't be accepted. Even if DrKonqi doesn't know any version information on bugs.kde.org, when it sends a request against some disabled version, bugs.kde.org (since upgrading to bugzilla 4.2.5) should detect and reject the request, then send back a error message to drkonqi, like the screenshot in https://bugs.kde.org/attachment.cgi?id=78600&action=edit.
Now this expected rejecting behavior works for versions like "4.9.0" or "4.10.60" (no whitespace), but not for "4.8.5 (4.8.5)" (containing whitespace). It might be that DrKonqi sends version containing whitespace in a wrong way, or that bugzilla deals with version containing white space in a wrong way. But since Drknonqi also sends distribution information (like "Ubuntu Packages") containing whitespace in the same way and bugzilla deals with that correctly, I'm inclined to think something goes wrong in bugzilla when dealing with versions.
- Jekyll
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://git.reviewboard.kde.org/r/110633/#review33110
-----------------------------------------------------------
On May 24, 2013, 2:54 p.m., Martin Gräßlin wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> http://git.reviewboard.kde.org/r/110633/
> -----------------------------------------------------------
>
> (Updated May 24, 2013, 2:54 p.m.)
>
>
> Review request for KDE Runtime, Plasma, Ben Cooksley, and Myriam Schweingruber.
>
>
> Description
> -------
>
> Bugzilla provides a feature to disable versions. This means that the developers do not want to have further reports for this version. Any crash report is by that not helpful any more. So let's just disable reporting crashes for such bugs.
>
> If this change gets accepted I intend to backport it to 4.10 and to inform kde-packagers about it to ship it as an update to *all* version they support. This would automatically prevent most duplicates report we get e.g. for KWin.
>
>
> Diffs
> -----
>
> drkonqi/reportinterface.cpp 4190c40
>
> Diff: http://git.reviewboard.kde.org/r/110633/diff/
>
>
> Testing
> -------
>
> See https://bugs.kde.org/show_bug.cgi?id=320217 - the bug was created with version 4.10.60. Afterward the version got disabled and DrKonqi doesn't allow me to report crashes for this version anymore.
>
>
> File Attachments
> ----------------
>
> With it disabled
> http://git.reviewboard.kde.org/media/uploaded/files/2013/05/24/drkonqi-disabled.png
>
>
> Thanks,
>
> Martin Gräßlin
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/plasma-devel/attachments/20130525/b5a74189/attachment.html>
More information about the Plasma-devel
mailing list