Re: SolarMutex: please educate me

On Tue, 2013-05-14 at 11:49 +0200, Stephan Bergmann wrote:
> On 05/13/2013 07:12 PM, Lionel Elie Mamane wrote:
> > Base or Base-related code seems to have some issues around the
> > "SolarMutex".
> >
> > https://bugs.freedesktop.org/show_bug.cgi?id=61725
> > https://bugs.freedesktop.org/show_bug.cgi?id=64241
> > (and others historically)
> >
> > Please educate me on the SolarMutex. What is its meaning, when should
> > be held/acquired, ... From
> > http://cgit.freedesktop.org/libreoffice/core/commit/?id=b34b8d3372364b3c5043da0357ec69505e8d8602,
> > my "guess" is that it is related to GUI stuff. Maybe it should be held
> > when calling a GUI (that is, GTK/Win32/QT/...) function? But it can't
> > be that simple, else vcl would "just do it"... It seems to expect
> > higher-level code to hold it already. Why?
>
> Historically, the SolarMutex was assumed to be a clever solution to the 
> problem of turning StarOffice from a single-threaded application to a 
> multi-threaded one---lock the SolarMutex around any code that is not yet 
> thread-safe, and be done with it.  Or so people thought.
>
> The world has long since found out that "single global mutex" is a 
> horrible idea for GUI applications (where all GUI frameworks have 
> settled on "confine relevant code to run in a single event-loop thread" 
> instead).
>
> This leaves us with completely unmaintainable ("where do I need to lock 
> SolarMutex?"---"nobody knows for sure") and broken (cf. 
> SolarMutexReleaser) code.
>
> Incidentally, I had exactly the same problem with 
> <https://gerrit.libreoffice.org/#/c/3894/> "TODO: Ensure SolarMutex is 
> locked at DBG_TESTSOLARMUTEX checkpoints" last night.

Yeah, and I'm continually dithering as to where to put one for the
various writer mail merge wizard "oh my god its not locked" asserts.

C.