Re: [dev] Problem with windows name
Thank you so much Страхиња Радић!
This works perfectly.
André.
On 10/07/2023 9:35, Страхиња Радић wrote:
> I managed to fix root window title being shown incorrectly by skipping
> conversion only for root window. I attached a patch with this message. Feel
> free to review it, I will make any necessary corrections.
> From c5b399e138f6873e2eb9a292d55c5382e654f074 Mon Sep 17 00:00:00 2001
> From: =?UTF-8?q?=D0=A1=D1=82=D1=80=D0=B0=D1=85=D0=B8=D1=9A=D0=B0=20=D0=A0?=
> =?UTF-8?q?=D0=B0=D0=B4=D0=B8=D1=9B?=
> Date: Mon, 10 Jul 2023 09:25:30 +0200
> Subject: [PATCH] dwm.c: Always convert title to UTF-8, except for root window
> MIME-Version: 1.0
> Content-Type: text/plain; charset=UTF-8
> Content-Transfer-Encoding: 8bit
>
> Window names containing only ISO 8859-1 were broken on the location of
> non-ASCII ISO 8859-1 character, for example áâãéíóôúç, because they were
> treated as UTF-8. When commenting out the if-branch testing for XA_STRING,
> root
> window title, usually set by status programs such as slstatus to UTF-8 text,
> was not shown correctly, so this patch adds an exception for root window
> instead, skipping conversion in that case.
>
> Reported by Dr. André Desgualdo Pereira.
>
> Signed-off-by: Страхиња Радић
> ---
> dwm.c | 5 -
> 1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/dwm.c b/dwm.c
> index f1d86b2..c896325 100644
> --- a/dwm.c
> +++ b/dwm.c
> @@ -917,7 +917,10 @@ gettextprop(Window w, Atom atom, char *text, unsigned
> int size)
> text[0] = '\0';
> if (!XGetTextProperty(dpy, w, , atom) || !name.nitems)
> return 0;
> - if (name.encoding == XA_STRING) {
> + /*if (name.encoding == XA_STRING) {
> + strncpy(text, (char *)name.value, size - 1);
> + } else */
> + if (w == DefaultRootWindow(dpy)) {
> strncpy(text, (char *)name.value, size - 1);
> } else if (XmbTextPropertyToTextList(dpy, , , ) >= Success
> && n > 0 && *list) {
> strncpy(text, *list, size - 1);
> --
> 2.38.5
>
--
Dr. André Desgualdo Pereira
Psiquiatra - CRM/SP: 120218 - RQE: 61032
WhatsApp: (11) 985-847-809 - email: des...@gmail.com
Consultório Lapa: Rua Clélia, 2208 - sala 307
Consultório Santana: Av. Gen. Ataliba Leonel, 93 - sala 61
Re: [dev] Problem with windows name
I managed to fix root window title being shown incorrectly by skipping
conversion only for root window. I attached a patch with this message. Feel
free to review it, I will make any necessary corrections.
From c5b399e138f6873e2eb9a292d55c5382e654f074 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=D0=A1=D1=82=D1=80=D0=B0=D1=85=D0=B8=D1=9A=D0=B0=20=D0=A0?=
=?UTF-8?q?=D0=B0=D0=B4=D0=B8=D1=9B?=
Date: Mon, 10 Jul 2023 09:25:30 +0200
Subject: [PATCH] dwm.c: Always convert title to UTF-8, except for root window
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Window names containing only ISO 8859-1 were broken on the location of
non-ASCII ISO 8859-1 character, for example áâãéíóôúç, because they were
treated as UTF-8. When commenting out the if-branch testing for XA_STRING, root
window title, usually set by status programs such as slstatus to UTF-8 text,
was not shown correctly, so this patch adds an exception for root window
instead, skipping conversion in that case.
Reported by Dr. André Desgualdo Pereira.
Signed-off-by: Страхиња Радић
---
dwm.c | 5 -
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/dwm.c b/dwm.c
index f1d86b2..c896325 100644
--- a/dwm.c
+++ b/dwm.c
@@ -917,7 +917,10 @@ gettextprop(Window w, Atom atom, char *text, unsigned int size)
text[0] = '\0';
if (!XGetTextProperty(dpy, w, , atom) || !name.nitems)
return 0;
- if (name.encoding == XA_STRING) {
+ /*if (name.encoding == XA_STRING) {
+ strncpy(text, (char *)name.value, size - 1);
+ } else */
+ if (w == DefaultRootWindow(dpy)) {
strncpy(text, (char *)name.value, size - 1);
} else if (XmbTextPropertyToTextList(dpy, , , ) >= Success && n > 0 && *list) {
strncpy(text, *list, size - 1);
--
2.38.5
signature.asc
Description: PGP signature
Re: [dev] Problem with windows name
Adding to gettextprop in dwm.c:
// ...
if (!XGetTextProperty(dpy, w, , atom) || !name.nitems)
return 0;
FILE* log = fopen("/home/user/dwm.log", "at");
fprintf(log, "---\n");
fprintf(log, "atom = {%lu}\n", atom);
fprintf(log, "text = {%s}\n", text);
fprintf(log, "value= {%s}\n", name.value);
fprintf(log, "encoding = {%lu}\n", name.encoding);
fprintf(log, "format = {%d}\n", name.format);
fprintf(log, "nitems = {%lu}\n", name.nitems);
fclose(log);
if (name.encoding == XA_STRING) {
// ...
I got:
---
atom = {39}
text = {}
value= {thistest.odt - LibreOffice Writer}
encoding = {31}
format = {8}
nitems = {34}
---
atom = {39}
text = {}
value= {thistestL.odt - LibreOffice Writer}
encoding = {385}
format = {8}
nitems = {38}
So COMPOUND_TEXT sets the encoding field of XTextProperty to 385, a value not
listed in Xatom.h, while STRING sets it to 31 (XA_STRING). The code:
if (name.encoding == XA_STRING) {
strncpy(text, (char *)name.value, size - 1);
} else if (XmbTextPropertyToTextList(dpy, , , ) >= Success
&& n > 0 && *list) {
strncpy(text, *list, size - 1);
XFreeStringList(list);
}
converts any string not having XA_STRING as encoding with
XmbTextPropertyToTextList[1].
The issue is that ISO 8859-1-encoded text, marked as XA_STRING, is copied with
strncpy, and treated as UTF-8 later on in drw_text in drw.c. It probably should
be converted to UTF-8 from the current locale, even when encoding field is set
to XA_STRING. However, when I commented out the first branch like this:
/*if (name.encoding == XA_STRING) {
strncpy(text, (char *)name.value, size - 1);
} else*/ if (XmbTextPropertyToTextList(dpy, , , ) >= Success
&& n > 0 && *list) {
strncpy(text, *list, size - 1);
XFreeStringList(list);
}
The document title in LibreOffice was shown correctly. *However*, the output
from slstatus, containing UTF-8 while having the encoding field also set to
XA_STRING, was *not* shown correctly, even though the description of
XmbTextPropertyToTextList states that
> The input text strings must be given in the current locale encoding (for
> XmbTextListToTextProperty and XwcTextListToTextProperty), or in UTF-8
> encoding
> (for Xutf8TextListToTextProperty).
and my LANG is set to sr_RS.UTF-8 (so my "current locale encoding" should be
UTF-8).
So, this would be a workaround, but one with further potential issues.
[1]: https://linux.die.net/man/3/xmbtextpropertytotextlist
Re: [dev] Problem with windows name
I debugged dwm, adding to drw.c:
static void
log_msg(const char* fmt, ...)
{
char buf[4096];
va_list args;
va_start(args, fmt);
vsnprintf(buf, sizeof(buf), fmt, args);
va_end(args);
fprintf(logfile, "%s\n", buf);
}
and calls to utf8decodebyte:
static long
utf8decodebyte(const char c, size_t *i)
{
for (*i = 0; *i < (UTF_SIZ + 1); ++(*i))
if (((unsigned char)c & utfmask[*i]) == utfbyte[*i])
{
log_msg("*i = %lu, for '%c' returning '%c'",
*i, c, (unsigned char)c & ~utfmask[*i]);
return (unsigned char)c & ~utfmask[*i];
}
return 0;
}
and drw_text:
utf8str = text;
nextfont = NULL;
while (*text) {
log_msg("*text == '0x%X' == '%c'",
*text, *text);
utf8charlen = utf8decode(text, ,
UTF_SIZ);
for (curfont = drw->fonts; curfont; curfont =
curfont->next) {
charexists = charexists ||
XftCharExists(drw->dpy, curfont->xfont, utf8codepoint);
I got the following output from "thisátest.odt"
// á
*text == '0xFFE1' == ''
*i = 3, for '' returning '^A'
*i = 1, for 't' returning 't'
*text == '0x74' == 't'
*i = 1, for 't' returning 't'
and the following from "thisátestњ.odt":
// á
*text == '0xFFC3' == ''
*i = 2, for '' returning '^C'
*i = 0, for '' returning '!'
[...]
// њ
*text == '0xFFD1' == ''
*i = 2, for '' returning '^Q'
*i = 0, for '<9A>' returning '^Z'
From here, it seems that dwm is receiving correct UTF-8 representations of "á"
(0xC3 0xA1) and "њ" (0xD1 0x9A) for "thisátestњ.odt", but it receives
ISO 8859-1 representation of "á" (no wonder, given it is passed a STRING
instead
of UTF8_STRING or COMPOUND_TEXT), 0xE1, followed by the next ASCII character,
0x74 ("t"), still interpreting the two as UTF-8 sequence, when those two bytes
form an invalid UTF-8. That invalid UTF-8 is further passed to libfreetype or
whatever, which just interrupts output at that point.
signature.asc
Description: PGP signature
Re: [dev] Problem with windows name
On 23/07/09 08:43AM, Dr. André Desgualdo Pereira wrote: > # xprop > xprop shows the correct name, but it doesn't has the _NET_WM_NAME defined as > it used to have on Debian 11 and before (I guess it is because missing > library > tkinter for python2.7). > Example: WM_NAME(STRING) = "André Desgualdo.odt - OpenOffice Writer" I have made two files, one with á and ASCII characters and another with added њ. The results from xprop are: WM_NAME(STRING) = "thisátest.odt - LibreOffice Writer" ^^ WM_NAME(COMPOUND_TEXT) = "thisátestњ.odt - LibreOffice Writer" ^ This implies that this is a X.Org issue. Strings having only ISO 8859-1 characters are silently degraded to STRINGs when they should really be COMPOUND_TEXT. signature.asc Description: PGP signature
[dev] Problem with windows name
# The Problem: dwm status bar fails to display the name of Libreoffice and OpenOffice windows # Conditions (when it happens): 1) Debian 12 (previous Debian worked fine) 2) The file name must contain a latin character (one of these: áâãéíóôúç) and NOT contain other glyphs characters (for example Japanese 你) # Examples: name of the filename displayed on dwm status "André Desgualdo.odt" -> "Andr Desgualdo.odt - OpenOffice Writer" (wrong, it miss the é) "Andre Desgualdo.odt" -> "Andre Desgualdo.odt - OpenOffice Writer" (right) "你 André Desgualdo.odt"-> "你 André Desgualdo.odt - OpenOffice Writer" (right) "André Desgualdo 你.odt"-> "André Desgualdo 你.odt - OpenOffice Writer" (right) # xprop xprop shows the correct name, but it doesn't has the _NET_WM_NAME defined as it used to have on Debian 11 and before (I guess it is because missing library tkinter for python2.7). Example: WM_NAME(STRING) = "André Desgualdo.odt - OpenOffice Writer" # xwininfo But xwininfo also fails to show the name properly. Example: xwininfo: Window id: 0xe031bd "Andr Desgualdo.odt - OpenOffice Writer" # Other window managers Other window managers show the window's name correctly. Tested: xfce4wm and i3wm. # Other programs that are not Libreoffice or OpenOffice When opening the same file with other program (for example opening "André Desgualdo.odt" with file-roller), the name is displayed correctly and xprop shows the following info: _NET_WM_NAME(STRING) = "Andr\303\251 Desgualdo.odt - OpenOffice"). # Workaround Any suggestion to further investigate or possible workarounds or fix would be really appreciated. (Currently I made a script to use xdotool to rename the window as soon as it is opened by OpenOffice).
