no border patch causes weird gimp interaction

[Zerogaku]

hi this is a pretty specific and unimportant issue but, when having gimp open on a separate empty tag and swapping to it, the tag displays the same content as the previous tag. Gimp is present just can't display correctly, only when closing and opening it again, or opening it on a tag that has 1 or more windows, this happens when a compositor is enabled for some reason, i've tried both xcompmgr and picom, the issue only goes away when no border patch is disabled. is there a fix for this?

gimp opens in tag 5:

swap to tag 5 to find gimp "invisible":

when trying to interact with the gimp window

[bakkeby]

Maybe.

Could you try with these changes?

diff --git a/dwm.c b/dwm.c
index 7497c65..20b9a7a 100644
--- a/dwm.c
+++ b/dwm.c
@@ -1455,6 +1455,35 @@ configure(Client *c)
        ce.width = c->w;
        ce.height = c->h;
        ce.border_width = c->bw;
+       #if NOBORDER_PATCH
+       if (((nexttiled(c->mon->clients) == c && !nexttiled(c->next))
+               #if MONOCLE_LAYOUT
+               || &monocle == c->mon->lt[c->mon->sellt]->arrange
+               #endif // MONOCLE_LAYOUT
+               #if DECK_LAYOUT
+               || (&deck == c->mon->lt[c->mon->sellt]->arrange &&
+                       c->mon->nmaster == 0)
+               #endif // DECK_LAYOUT
+               #if FLEXTILE_DELUXE_LAYOUT
+               || (&flextile == c->mon->lt[c->mon->sellt]->arrange && (
+                       (c->mon->ltaxis[LAYOUT] == NO_SPLIT &&
+                        c->mon->ltaxis[MASTER] == MONOCLE) ||
+                       (c->mon->ltaxis[STACK] == MONOCLE &&
+                        c->mon->nmaster == 0)))
+               #endif //FLEXTILE_DELUXE_LAYOUT
+               )
+               #if FAKEFULLSCREEN_CLIENT_PATCH && !FAKEFULLSCREEN_PATCH
+               && (c->fakefullscreen == 1 || !c->isfullscreen)
+               #else
+               && !c->isfullscreen
+               #endif // FAKEFULLSCREEN_CLIENT_PATCH
+               && !c->isfloating
+               && c->mon->lt[c->mon->sellt]->arrange) {
+               ce.width += c->bw * 2;
+               ce.height += c->bw * 2;
+               ce.border_width = 0;
+       }
+       #endif // NOBORDER_PATCH
        ce.above = None;
        ce.override_redirect = False;
        XSendEvent(dpy, c->win, False, StructureNotifyMask, (XEvent *)&ce);

The idea is that the same changes that are applied in resizeclient must also be applied in configure.

[Zerogaku]

Thanks adding that fixed it. Krita also seemed to have that very same issue and it got fixed as well (i guess for any apps that show a splash screen first) just curious what does the function configure do?

[Zerogaku]

Ah i've noticed a small side affect, it trims the edges of the windows

krita:

Gimp:

Emacs:

the fix seems to be switching to another tag and switching back (doesn't work all the time)

this is how i positioned the code:

	ce.x = c->x;
	ce.y = c->y;
	ce.width = c->w;
	ce.height = c->h;
	ce.border_width = c->bw;
        ce.height = c->h;
	ce.border_width = c->bw;
        if (((nexttiled(c->mon->clients) == c && !nexttiled(c->next))
                || &monocle == c->mon->lt[c->mon->sellt]->arrange
                || (&deck == c->mon->lt[c->mon->sellt]->arrange &&
                        c->mon->nmaster == 0)
                )
                && !c->isfullscreen
                && !c->isfloating
                && c->mon->lt[c->mon->sellt]->arrange) {
                ce.width += c->bw * 2;
                ce.height += c->bw * 2;
                ce.border_width = 0;
        }
	ce.above = None;
	ce.override_redirect = False;
	XSendEvent(dpy, c->win, False, StructureNotifyMask, (XEvent *)&ce);

[bakkeby]

Hi @Zerogaku,

can you try removing these two lines:

                ce.width += c->bw * 2;
                ce.height += c->bw * 2;

I see that the border width is already added to c->w and c->h in the resizeclient function.

[Zerogaku]

Hi @Zerogaku,

can you try removing these two lines:

                ce.width += c->bw * 2;
                ce.height += c->bw * 2;

I see that the border width is already added to c->w and c->h in the resizeclient function.

unfortunately this causes the initial problem to return

[bakkeby]

OK, thanks for checking. Sounds like it may be a picom issue, what version are you using?

[Zerogaku]

the picom version is vgit-751f3 which i believe is the latest version, but this issue also occurs with xcompmgr so this might be a generic compositor issue or dwm

[bakkeby]

To be fair the noborder patch is notorious for creating weird side effects.