freenode/#clim - IRC Chatlog
Search
7:20:47
jackdaniel
scymtym: looking at the branch you've pointed me to: I'm not certain which changes are cleaned up; if I had to guess that would be two commits 4f125a2e and 0eec2e05; are changes after that also something we need?
7:26:20
jackdaniel
(n.b in your branch I have unbalanced enter events for randomized sheet hierarchy
7:28:26
scymtym
jackdaniel: it is not finished. it's just what i have at the moment. since we agreed you would have a look some time ago
7:29:52
scymtym
i was improving things whack-a-mole-style until you noticed the menu problem. since that may require more changes, i stopped ironing out the enter/exit problems
7:30:33
scymtym
did you figure out how to use the event logging? i used it to debug event synthesis and dispatching
7:35:07
jackdaniel
I'll look around how you add highlight and if I think of something better I'll let you know
7:57:26
scymtym
i missed your question earlier. bbc8ca822 has more fixes. but i didn't clean up or comment those since i expect more changes to get menus working
7:58:30
scymtym
no, missing highlighting doesn't indicate a problem. but the highlighting sometimes catches problems like wrong port pointer sheet or unbalanced events
8:00:05
scymtym
that's highlighting. by logging i mean the inspector that opens when you load event-log.lisp. the forms at the bottom of the file control it
8:01:11
jackdaniel
hilighting works only per the around method you have defined in the hierarchy-tool (i.e patterns from clouseau sheets.lisp does not appear, I'm not sure how to enable this kind of highlightning)
8:02:06
jackdaniel
also, it highlights only panes, but that is understandable, because this protocol is pane-specific
8:02:51
scymtym
yeah, too many kinds of highlighting. moving the pointer over events or sheets in the inspector should also highlight sheets (with yellow stripes). that works best for single-mirror mode, of course
8:03:47
jackdaniel
when I hover the presentation in the inspectorw window I see the pattern on the sheet
8:05:13
jackdaniel
so, could you reiterate what is the problem you've encountered? (sorry for being so slow :)
8:07:30
scymtym
there are at least three: 1) menus are broken (with and without my additional changes afaict) 2) you found unbalanced enter/exit events for the random mirroring case 3) there sometimes are unbalanced enter/exit events for non-leaf sheets
8:09:13
jackdaniel
menus as in the issue I've commented on the PR? regarding enter/exit events in 2) and 3) I think it is somewhat a regression from what is in ieh-1 branch, no?
8:10:30
jackdaniel
the inspector highlight tool does not work perfectly (i.e it sometimes wipes out whole window without repainting it, not all events provide highlight), could it be that I do something wrong?
8:15:59
scymtym
sure, both debugging tools are not perfect, but at least the event log itself should be reliable
8:17:50
jackdaniel
try ieh-2 branch, unless you drag-and-drop between two mirrored sheets (spurious ungrab events I've mentioned), all is balanced at least from perspective of the counter in hierarchy-tool demo
8:18:02
scymtym
if i remember correctly, ieh-1 had a problem with enter/exit events when pressing a button over a sheet, then holding and moving the pointer somewhere else and releasing there
8:18:44
jackdaniel
yes, that's the only fail scenario, while on the branch you've shown even without pressing I have unbalanced events in random hierarchy
8:19:02
scymtym
the counter is only visible for "leaf" sheets. top-level and {h,v}rack panes were never properly balanced, i think
8:19:56
jackdaniel
either way, I'll start from tackling menu issue and then we'll pick it up from there
8:22:18
jackdaniel
I'd prefer to solve it by making menu use click-to-focus, but that is not ready, so I'll probably try something in line with: menu root when ungrabbed determines which menu element was pressed. but this is very vague, I didn't look into the code yet
8:23:17
jackdaniel
arming/disarming seems to work fine what may be a clue for a correct solution to this
8:24:41
scymtym
the problem is that two release events are distributed due to grabbing. the menu root gets its event first, the menu leaf gets the second one
8:25:12
scymtym
but the menu root disarms and tears down everything so the second event has no effect
8:25:50
scymtym
it could be "fixed" by dispatching grabbing related events in the opposite order, but that would be a fragile hack
8:28:27
jackdaniel
extending menu demo shown another regression with sheet-is-not-ancestor event signalled
8:29:08
scymtym
also, yes, in the long run i would also prefer to change how menus work. but for now we should restore the event system to a state where it can support the current menu behavior
8:30:05
scymtym
this is getting overwhelming. after sorting out the event issues, there two probably very deep clx bugs to figure out
8:34:05
jackdaniel
could you merge this if you find no problems in the commit? I don't want to obfuscate changes in the branch
8:39:31
jackdaniel
you have to forgive me poor jokes relating memes and anime (or, given I want to look less silly, relating to movies and japanease language ,)
8:43:09
scymtym
i can tolerate it. i might retaliate with "they set us up the bomb", "for great justice" or even "get in the robot, shinji" jokes, though (i'm old). i have a suspicion that ck_ is sitting on a few of those as well
8:45:10
scymtym
but eastereggs and jokes require delivering something solid first, so we have our work cut out for us
9:12:20
jackdaniel
n.b: it doesn't seem to be related to port-grabbed-sheet handling, when grabbed-sheet hijack is disabled problem still occurs
9:14:40
jackdaniel
event is sent twice from the xserver (because of grabbing), submenus have separate frames