On Mon, Apr 25, 2022 at 05:38:35PM +0100, Chris Down wrote:
> Hiltjo Posthuma writes:
> > Whats the similar issue exactly? Does this issue also happen when bisecting the
> > same commit (so its a also regression)?
>
> For 8806b6e23793 ("manage: propertynotify: Reduce cost of unused size
> hints"), the issue is that c->isfixed may not be set early, which might
> affect whether we intend to make the window floating or not.
>
> For bece862a0fc4 ("manage: For isfloating/oldstate check/set, ensure trans
> client actually exists"), the issue is that some non-SDL client which
> actually _should_ float have WM_TRANSIENT_FOR set to the root window. I
> posted about this on hackers_AT_ a few weeks ago.[0]
>
> For 8806b6e23793, I think the change I suggested should be sufficient, but
> would like to get confirmation from Ethan first and then will submit it as a
> proper patch.
>
Yes please
> For bece862a0fc4, I think it might just need to be reverted unless more
> specific logic can be applied, since it seems over the top to add more logic
> just to reliably distinguish SDL clients from other affected clients.
>
I've reverted the patch. It can be reworked to fix both issues (the regression and the initial
reason it was posted) maybe.
> In both cases the symptoms are that some subset of clients may not float
> when they should, but in the case of 8806b6e23793 it's "fixed" windows where
> the dimensions are fixed in size hints, and in the case of bece862a0fc4 it's
> windows with WM_TRANSIENT_FOR set to the root window (like the gpg2
> pinentry).
>
> 0: https://lists.suckless.org/hackers/2203/18220.html
--
Kind regards,
Hiltjo
Received on Tue Apr 26 2022 - 10:50:25 CEST