GraphicsMagick: add debug variant

[mascguy (Christopher Nielsen)]

Presently, we're trying to diagnose a crash occurring for darktable, which appears to be originating from within this port:

issue:64252

There are two barriers, though:

• The GraphicsMagick library installs a SIGSEGV handler, preventing the app from catching/handling the condition;
• No +debug variant, which would assist with diagnostics.

For the debug variant, that should be reasonably straightforward, and would be beneficial.

[mascguy (Christopher Nielsen)]

Too much on my plate, I'll let Ryan tackle this

[mascguy (Christopher Nielsen)]

issue:64252 has been resolved, so this is no longer needed.

If there's user demand, we can always reopen.

[jmroot (Joshua Root)]

There is still the valid issue raised in the other ticket that GraphicsMagick installs its own SIGSEGV handler when used as a library. Up to you if you want to open a ticket for that, here or upstream.

[mascguy (Christopher Nielsen)]

Let's expand the scope of this ticket to include the lib signal handler, per @jmroot in comment:3

[ryandesign (Ryan Carsten Schmidt)]

Replying to mascguy:

Relative to the signal handler, we should consider patching the code to disable that, for the library. Along with engaging with upstream, for a longer-term fix.

Yes, please ​consult with upstream before attempting that.

[mascguy (Christopher Nielsen)]

Relative to the signal handlers, it looks like apps can now explicitly disable those. Functionality was added to the GraphicsMagick API in version 1.3.35, released 2020-02-23:

InitializeMagickEx(): New function which may be used in place of InitializeMagick() to initialize GraphicsMagick. This initialization function returns an error status value, may update a passed ExceptionInfo structure with error information, and provides an options parameter which supports simple bit-flags to tailor initialization. The signal handler registrations are skipped if the MAGICK_OPT_NO_SIGNAL_HANDER flag is set in the options.

@parafin, is that something you folks want to try, for darktable?

[mascguy (Christopher Nielsen)]

Since apps now have control over signal handler registration, we can take that off of our plate.

As for the debug variant: Based on local testing, the debug portgroup does the trick for this port. Are there any concerns/gotchas with utilizing that?