On 11/27/22 18:57, Miller Puckette via Pd-dev wrote:
TO pd dev -
I'm fixing to make a bugfix release. I just merged 'develop' and 'Documentation'. Anything else (fixes only, please, no new features) that I should look at?
i'm wondering about the major iemgui regression. as of now, any external library that uses the g_all_guis.h infrastructure (e.g. the "iemgui" library, or moonlib's mknob) crashes with Pd-0.53. antoine has updated moonlib to use the new infrastructure (but the binaries are now Pd-0.53 only). i started updating the "iemgui" library (but i haven't completed that yet).
so i wonder whether: - we should keep binary compatibility with old binaries, probably at the cost of keeping two distinct iemgui-frameworks in the sources (g_all_guis-legacy and g_all_guis-new) - we should retire the public iemgui API (just like we did for "error()"), so the only API for creating GUI externals is directly via the widgetbehaviour functions.
the two questions are somewhat orthogonal, but i think the question ought to be tackled.
as i've said before, my personal perference would be at retracting the API and not caring about ABI breakage, but that is a very radical approach, and rather unusual for the Pd world.
i am not very optimistic about being able to merge the new-style and legacy g_all_guis into a single ABI-compatible (with older binaries) interface. hence my radical approach. but i can totally understang if such an approach is not acceptable.
i also do not have any numbers, on how many externals are actually affected by the current breakage (i only know of moonlib and iemgui; but there might be others)
gfmdsa IOhannes