GitLab

*updated most of the audio API to the latest version (with the exception of ALSA and JACK which are partial until our improvements can be merged and/or new implementation of JACK thoroughly tested to ensure we like its default behavior)

*removed text editor since now pd-l2ork's facility for editing text inside object boxes is adequate enough not to require a full-fledged editor

Patch put on tracker for Pd-extended, but I thought I'd go ahead and commit here, too

This change allows gobjs to be pointed to by gpointers.  This makes it
possible to improve the interface for the *info objects, as they can pass
around lists of gpointers to refer to objects/canvases.  This is better than
using index numbers and canvas levels, which is cumbersome and in some cases
much slower (since a walk through the glist is required each time to find
the referrent for an index).

* properly escaped "boxtext" output
* output expanded args for canvasinfo "args"

*added new files from the latest pure-data/externals/maxlib tree to git (missing files resultted in a failed build)

*fixed regression where cnv objects captured runtime clicks due to pd-l2ork's passing of the click to the topmost visible object, rather than the first clickable object in the stack. This also fixes problems with legacy patches that rely vanilla/extended's assignment of click to the lowest, and as is the case with s-env-help.pd in s-abstractions, invisible object, rather than the one apparent to the user.

*finally fixed the memory leak and improved how the newlines are created inside objects (needs further testing)

*fixed positioning of autopatched objects that were placed immediately below typed objects that have not been activated yet.

*fixed inconsistent display of array styles between pd-l2ork and vanilla. LATER consider cleaning this up by adjusting g_canvas.h instead of all these workarounds.

*backported vanilla patch http://sourceforge.net/p/pure-data/pure-data/ci/f535c339c99a74edb3920c2e630500df128b7d8f/

fixed bug with 8x8 click "hotspot" not working correctly on x-axis of garrays

*added sanity check for mknob (a subset of iemgui) to see if selected status would require object to be highlighted (if inside graph-on-parent, it should be silently selected but not visibly selected)
*fixed regression where selected variable in the old_g_all_guis.inc did not allow for a value larger than 1-bit, resulting in selected=2 being selected=0

*TODO: why is alsa's snd_ctl_card_info_free(info) causing double free error (valgrind's "Adress  is is 40 bytes inside a block of size 376 free'd)? No clue. But with that call disabled, everything works as it should and valgrind reports no further errors.

*fixed preset segfault when doing clear on a hubless node
*clean-up

*made startmotion more user-friendly by hiding iemgui's handles after initial creation until motion is complete. LATER get rid of this because it is likely to be seen as an interim hack due to determinacy issues between the gui and core engine.

*more thorough fix for the preset_node's second object getting incorrectly invisibly connected back to the preset_hub

*made K12 object creation more robust (removing susceptibility to relative paths from the current location)
*ensured that the ctrl key is released when opening various prompts (there may be possibly others)