On 28 July 2015 at 11:31, Alexandre Bergel <alexandre.bergel@me.com> wrote:
Hi!
Related to this, I find very painful to write unit tests for gui. Generating event by hand is quite tedious. Someone has worked on that direction already?
Yeah, i spent an hour playing with making benchmark on handling mouse move events.. Had quite random results, not giving any clue where the bottleneck/problem is. Decrypting & interpreting message tally is hell of a pain. Especially when it is not quite capable to show all stats about where time spent (including simple sleeping time)... or maybe because i am incapable to interpret it correctly.. which at the end gives same result: no result :)
Cheers Alexandre
Le 27 juil. 2015 à 10:42, Igor Stasenko <siguctua@gmail.com> a écrit :
Nothing here is about things to fix, but rather about how impossible to debug event handling in system that runs and relies on very same events.. I had very annoying problem with mouse move handling and processing it correctly, and i finally solved the problem after number of attempts to track down where problem lies in..
So, here is small list of problems i faced:
- broken image (image won't start up) , because i send unimplemented message in critical place, thus image simply non-responsible/crashing immediately after startup. Easy to fix (sarcasm)
- debugger window spam on every event (because of DNU/other exception). A slightly friendlier problem than above :)
- and last one, if you break some event handling/delivery, even if image "behaves" you simply left without events doing what is needed, means you got another form of a broken image, which you cannot use anymore and need to start over and revert the changes you made.
- veeeerrrrryyy slow event handling, because of printing every event into transcript , trying to see where the shit lies.
- decrypting kilometers long event traces to see what goes where
But when things finally start work properly, you feel excited :)
-- Best regards, Igor Stasenko.
-- Best regards, Igor Stasenko.