--headless broken with --enable-headless

Stephan Bergmann sbergman at redhat.com
Wed Dec 12 08:19:18 PST 2012

On 12/12/2012 04:50 PM, Michael Meeks wrote:
> On Wed, 2012-12-12 at 15:44 +0100, Stephan Bergmann wrote:
>> Now, on yet closer inspection, it is only the versions of
>> CreateSalInstance for Android and --enable-headless
>> (vcl/android/androidinst.cxx, vcl/headless/headlessinst.cxx, both based
>> on SvpSalInstance) that do not lock the solar mutex
> 	Ah indeed - digging deeper you're right. It's rather non-obvious
> though. I'd still rather prefer to have it locked in one place centrally
> I guess, but ... it'll do for now.

Indeed; further clean-up most welcome.

> As such the extra lock/unlock around
> event emission shouldn't be necessary in either headless or gtk+ AFAICS.

With the working hypothesis now being that Application::Yield shall 
always be called with the solar mutex locked, then yes, those should not 
be necessary.  However, with VCL in general, and Application::Yield 
apparently not only called directly from Application::Execute in 
particular, I wouldn't dare to decide whether that hypothesis actually 
matches reality.


More information about the LibreOffice mailing list