diff options
author | Tomas Janousek <tomi@nomi.cz> | 2021-05-14 21:47:41 +0100 |
---|---|---|
committer | Tomas Janousek <tomi@nomi.cz> | 2021-05-14 22:05:09 +0100 |
commit | f8c835a33a7acd6f60e27f82ac40027d3ae15e25 (patch) | |
tree | 2a61095406d166d0803c3b51dc4ac540e67fd80a /bench/main.hs | |
parent | 141f51828df91aca5508902164e7bbaa691ae759 (diff) | |
download | xmobar-f8c835a33a7acd6f60e27f82ac40027d3ae15e25.tar.gz xmobar-f8c835a33a7acd6f60e27f82ac40027d3ae15e25.tar.bz2 |
Fix delayed reaction to USR1/2 signals
While using xmobar to reproduce/fix a bug in xmonad I noticed that
xmobar doesn't react immediately to the signals to reposition or change
screen.
Apparently none of the Xlib functions called from `repositionWin` flush
the Xlib output buffer (XMoveResizeWindow is known to not flush,
although it's unfortunately not documented), so when compiled with the
threaded runtime that runs XNextEvent in a separate thread, the
reposition is sent to the X server only later when other stuff happens.
With all monitors set to refresh once a minute, this can take a looong
time.
(It's entirely possible this happens even without the threaded runtime,
I didn't test that, sorry.)
The fix is to call XSync at the end of `repositionWin`.
While at it, I spotted that drawInWin calls XSync with discard=true,
which seems like a mistake. We don't want to discard any events, do we?
(In practice, I'd expect the `eventer` thread to read all events even
before the drawing thread manages to discard them, so even if this
discarding XSync ever had any reason to be there, I don't think it does
anything meaningful these days. I may be wrong, though.)
Diffstat (limited to 'bench/main.hs')
0 files changed, 0 insertions, 0 deletions