launchpad-dev team mailing list archive

[Martin Pool <mbp@xxxxxxxxxxxxx>]

On 10 January 2011 07:28, Benji York <benji.york@xxxxxxxxxxxxx> wrote:
> On Mon, Jan 10, 2011 at 7:48 AM, Julian Edwards
> <julian.edwards@xxxxxxxxxxxxx> wrote:
>> So, for some reason that I can't quite fathom, but it's something to do with
>> zcml, lp.shipit is getting imported, which in turn seems to import the whole
>> world.
>>
>> The whole world includes lp.testing, which in turn imports launchpadlib and
>> hence the keyring.  Which installs a SIGCHLD handler.
>>
>> Ummm, why is lp.shitpit doing that?
>
> It's not lp.shipit's fault, it's a stupidity in the keyring package.  It
> (keyring) does work at import time (an almost universal no-no) by
> initializing it's selected back-end (in your case KWallet).  I have a
> launchpadlib branch that I'll be landing soonish (and then a release,
> and then a patch to LP proper to use the new launchpadlib) which will
> delay the import of keyring until needed.  After that I'll see if
> there's anything I can do about keyring doing the import-time work and
> why it's handling SIGCHLD.
>
> I /think/ only people with KDE installed are affected.  So I'm inclined
> to suggest patching your local LP to use the pre-keyring version of
> launchpadlib (1.7.0).  If that's not enough, we can roll back to the
> pre-keyring version of launchpadlib.
>
> I expect to have this fixed within the next 9 hours.

It may be deciding that it's running under KDE by looking at the
environment.  I suspect if you unset KDE_FULL_SESSION in the terminal
where you're going to run the tests, it will fix it.

-- 
Martin