Hello,
while testing my xinerama re-implementation I noticed that x2goclient reports a wrong width sometimes when simply moving the window without changing the size. The reported size differs by one pixel (too much or too little). When running x2goclient --debug you can see this:
x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 329, y: 214, w: 1440, h: 900" x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 365, y: 213, w: 1440, h: 900" x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 825, y: 190, w: 1440, h: 900" x2go-DEBUG-../src/sshprocess.cpp:189> executing remote command via SshProcess object (24): "export DISPLAY=:50;printf '\''1095 0 345 900\n0 0 1095 900'\'' > $HOME/.x2go/C-uli-50-1431636769_stDKDE_dp24/xinerama.conf"
-> here the 345 should have been a 344.
Interestingly my patched x2goagent (that completely ignores xinerama.conf) sees that value. So I conclude that x2goclient is manipulating the nxproxy window size by one pixel itself. Disabling xinerama in the session's configuration fixes the problem.
That off-by-one behaviour is caused by this code in onmainwindow.cpp:
void ONMainWindow::slotXineramaConfigured() { if (resumingSession.fullscreen) return; if (xinSizeInc == -1) xinSizeInc=1; else xinSizeInc=-1; #ifdef Q_OS_LINUX lastDisplayGeometry.setWidth(lastDisplayGeometry.width()+xinSizeInc); XSync(QX11Info::display(),false); XResizeWindow(QX11Info::display(), proxyWinId, lastDisplayGeometry.width(),lastDisplayGeometry.height()); XSync(QX11Info::display(),false); #endif #ifdef Q_OS_WIN QRect geom; wapiWindowRect ( (HWND) proxyWinId, geom ); wapiMoveWindow( (HWND) proxyWinId, geom.x(), geom.y(), geom.width()+xinSizeInc, geom.height(),true); lastDisplayGeometry=proxyWinGeometry(); #endif xineramaTimer->start(500); }
If I set xinSizeInc to zero in all cases the off-by-one behaviour is completely gone. Why does xinSizeInc exist at all?
Uli
Package: x2goclient Severity: important Version: 4.0.4.0
Hi Uli, hi Alex,
I turn this discussion into a bug report against X2Go Client.
Short summary for Alex: Ulrich has completely revisited [2] Xinerama
in nx-libs recently for the 3.6.x branch of nx-libs [1].
The complete Xinerama code has been rewritten and moved mainly into
the hw/nxagent/Screen.c file, so that we can soon drop your patched
libNX_Xinerama.so library and all that LD_LIBRARY_PATH stuff from
x2goruncommand.
With Ulrichs work, the hacky part in X2Go Client of writing and
continuously updating a xinerama.conf file server-side from the
client-side becomes obsolete and should only be kepts as legacy code
for cases where X2Go Client talks to nxagent servers << 3.6.
I'd also appreciate it if you could give feedback on the below question.
Thanks, Mike
[1] https://github.com/ArcticaProject/nx-libs [2] https://github.com/ArcticaProject/nx-libs/pull/23
On Do 14 Mai 2015 23:32:03 CEST, Ulrich Sibiller wrote:
while testing my xinerama re-implementation I noticed that x2goclient reports a wrong width sometimes when simply moving the window without changing the size. The reported size differs by one pixel (too much or too little). When running x2goclient --debug you can see this:
x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 329, y: 214, w: 1440, h: 900" x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 365, y: 213, w: 1440, h: 900" x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 825, y: 190, w: 1440, h: 900" x2go-DEBUG-../src/sshprocess.cpp:189> executing remote command via SshProcess object (24): "export DISPLAY=:50;printf '\''1095 0 345 900\n0 0 1095 900'\'' > $HOME/.x2go/C-uli-50-1431636769_stDKDE_dp24/xinerama.conf"
-> here the 345 should have been a 344.
Interestingly my patched x2goagent (that completely ignores xinerama.conf) sees that value. So I conclude that x2goclient is manipulating the nxproxy window size by one pixel itself. Disabling xinerama in the session's configuration fixes the problem.
That off-by-one behaviour is caused by this code in onmainwindow.cpp:
void ONMainWindow::slotXineramaConfigured() { if (resumingSession.fullscreen) return; if (xinSizeInc == -1) xinSizeInc=1; else xinSizeInc=-1; #ifdef Q_OS_LINUX lastDisplayGeometry.setWidth(lastDisplayGeometry.width()+xinSizeInc); XSync(QX11Info::display(),false); XResizeWindow(QX11Info::display(), proxyWinId, lastDisplayGeometry.width(),lastDisplayGeometry.height()); XSync(QX11Info::display(),false); #endif #ifdef Q_OS_WIN QRect geom; wapiWindowRect ( (HWND) proxyWinId, geom ); wapiMoveWindow( (HWND) proxyWinId, geom.x(), geom.y(), geom.width()+xinSizeInc, geom.height(),true); lastDisplayGeometry=proxyWinGeometry(); #endif xineramaTimer->start(500); }
If I set xinSizeInc to zero in all cases the off-by-one behaviour is completely gone. Why does xinSizeInc exist at all?
--
DAS-NETZWERKTEAM mike gabriel, herweg 7, 24357 fleckeby fon: +49 (1520) 1976 148
GnuPG Key ID 0x25771B31 mail: mike.gabriel@das-netzwerkteam.de, http://das-netzwerkteam.de
freeBusy: https://mail.das-netzwerkteam.de/freebusy/m.gabriel%40das-netzwerkteam.de.xf...
Hi all,
It's a great news. Do we need to make a changes in the X2Go Client for this? I think it should be sufficient to make changes in x2goruncommand. Should user still have possibility to turn Xinerama off? I can still imagine the use case when user have an x2goagent window in the middle of two physical monitors but want to see agent window as one display.
regards Alex
Am 15.05.2015 um 13:59 schrieb Mike Gabriel:
Package: x2goclient Severity: important Version: 4.0.4.0
Hi Uli, hi Alex,
I turn this discussion into a bug report against X2Go Client.
Short summary for Alex: Ulrich has completely revisited [2] Xinerama in nx-libs recently for the 3.6.x branch of nx-libs [1].
The complete Xinerama code has been rewritten and moved mainly into the hw/nxagent/Screen.c file, so that we can soon drop your patched libNX_Xinerama.so library and all that LD_LIBRARY_PATH stuff from x2goruncommand.
With Ulrichs work, the hacky part in X2Go Client of writing and continuously updating a xinerama.conf file server-side from the client-side becomes obsolete and should only be kepts as legacy code for cases where X2Go Client talks to nxagent servers << 3.6.
I'd also appreciate it if you could give feedback on the below question.
Thanks, Mike
[1] https://github.com/ArcticaProject/nx-libs [2] https://github.com/ArcticaProject/nx-libs/pull/23
On Do 14 Mai 2015 23:32:03 CEST, Ulrich Sibiller wrote:
while testing my xinerama re-implementation I noticed that x2goclient reports a wrong width sometimes when simply moving the window without changing the size. The reported size differs by one pixel (too much or too little). When running x2goclient --debug you can see this:
x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 329, y: 214, w: 1440, h: 900" x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 365, y: 213, w: 1440, h: 900" x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 825, y: 190, w: 1440, h: 900" x2go-DEBUG-../src/sshprocess.cpp:189> executing remote command via SshProcess object (24): "export DISPLAY=:50;printf '\''1095 0 345 900\n0 0 1095 900'\'' > $HOME/.x2go/C-uli-50-1431636769_stDKDE_dp24/xinerama.conf"
-> here the 345 should have been a 344.
Interestingly my patched x2goagent (that completely ignores xinerama.conf) sees that value. So I conclude that x2goclient is manipulating the nxproxy window size by one pixel itself. Disabling xinerama in the session's configuration fixes the problem.
That off-by-one behaviour is caused by this code in onmainwindow.cpp:
void ONMainWindow::slotXineramaConfigured() { if (resumingSession.fullscreen) return; if (xinSizeInc == -1) xinSizeInc=1; else xinSizeInc=-1; #ifdef Q_OS_LINUX lastDisplayGeometry.setWidth(lastDisplayGeometry.width()+xinSizeInc); XSync(QX11Info::display(),false); XResizeWindow(QX11Info::display(), proxyWinId, lastDisplayGeometry.width(),lastDisplayGeometry.height()); XSync(QX11Info::display(),false); #endif #ifdef Q_OS_WIN QRect geom; wapiWindowRect ( (HWND) proxyWinId, geom ); wapiMoveWindow( (HWND) proxyWinId, geom.x(), geom.y(), geom.width()+xinSizeInc, geom.height(),true); lastDisplayGeometry=proxyWinGeometry(); #endif xineramaTimer->start(500); }
If I set xinSizeInc to zero in all cases the off-by-one behaviour is completely gone. Why does xinSizeInc exist at all?
x2go-dev mailing list x2go-dev@lists.x2go.org http://lists.x2go.org/listinfo/x2go-dev
--
Oleksandr Shneyder | Email: o.shneyder@phoca-gmbh.de phoca GmbH | Tel. : 0911 - 14870374 0 Ludwig-Feuerbach-str. 18 | Fax. : 0911 - 14870374 9 D-90489 Nürnberg | Mobil: 0163 - 49 64 461
Geschäftsführung: Dipl.-Inf. Oleksandr Shneyder
Amtsgericht München | http://www.phoca-gmbh.de HRB 196 658 | http://www.x2go.org USt-IdNr.: DE281977973
Hi Alex, (re-adding #872 to Cc:, dropping x2go-dev instead)
(my answers in-line at end of mail text)
On So 17 Mai 2015 08:45:57 CEST, Oleksandr Shneyder wrote:
Am 15.05.2015 um 13:59 schrieb Mike Gabriel:
Package: x2goclient Severity: important Version: 4.0.4.0
Hi Uli, hi Alex,
I turn this discussion into a bug report against X2Go Client.
Short summary for Alex: Ulrich has completely revisited [2] Xinerama in nx-libs recently for the 3.6.x branch of nx-libs [1].
The complete Xinerama code has been rewritten and moved mainly into the hw/nxagent/Screen.c file, so that we can soon drop your patched libNX_Xinerama.so library and all that LD_LIBRARY_PATH stuff from x2goruncommand.
With Ulrichs work, the hacky part in X2Go Client of writing and continuously updating a xinerama.conf file server-side from the client-side becomes obsolete and should only be kepts as legacy code for cases where X2Go Client talks to nxagent servers << 3.6.
I'd also appreciate it if you could give feedback on the below question.
Thanks, Mike
[1] https://github.com/ArcticaProject/nx-libs [2] https://github.com/ArcticaProject/nx-libs/pull/23
On Do 14 Mai 2015 23:32:03 CEST, Ulrich Sibiller wrote:
while testing my xinerama re-implementation I noticed that x2goclient reports a wrong width sometimes when simply moving the window without changing the size. The reported size differs by one pixel (too much or too little). When running x2goclient --debug you can see this:
x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 329, y: 214, w: 1440, h: 900" x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 365, y: 213, w: 1440, h: 900" x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 825, y: 190, w: 1440, h: 900" x2go-DEBUG-../src/sshprocess.cpp:189> executing remote command via SshProcess object (24): "export DISPLAY=:50;printf '\''1095 0 345 900\n0 0 1095 900'\'' > $HOME/.x2go/C-uli-50-1431636769_stDKDE_dp24/xinerama.conf"
-> here the 345 should have been a 344.
Interestingly my patched x2goagent (that completely ignores xinerama.conf) sees that value. So I conclude that x2goclient is manipulating the nxproxy window size by one pixel itself. Disabling xinerama in the session's configuration fixes the problem.
That off-by-one behaviour is caused by this code in onmainwindow.cpp:
void ONMainWindow::slotXineramaConfigured() { if (resumingSession.fullscreen) return; if (xinSizeInc == -1) xinSizeInc=1; else xinSizeInc=-1; #ifdef Q_OS_LINUX lastDisplayGeometry.setWidth(lastDisplayGeometry.width()+xinSizeInc); XSync(QX11Info::display(),false); XResizeWindow(QX11Info::display(), proxyWinId,
lastDisplayGeometry.width(),lastDisplayGeometry.height()); XSync(QX11Info::display(),false); #endif #ifdef Q_OS_WIN QRect geom; wapiWindowRect ( (HWND) proxyWinId, geom ); wapiMoveWindow( (HWND) proxyWinId, geom.x(), geom.y(), geom.width()+xinSizeInc, geom.height(),true); lastDisplayGeometry=proxyWinGeometry(); #endif xineramaTimer->start(500); }
If I set xinSizeInc to zero in all cases the off-by-one behaviour is completely gone. Why does xinSizeInc exist at all?
It's a great news. Do we need to make a changes in the X2Go Client for
X2Go Client needs to test the remote x2goagent version and check
whether it is << 3.5.99.0 or >= 3.5.99.0. If it is >= 3.5.99.0, the
new Xinerama support will be present (once we have released 3.5.99.0,
at the moment the code has not been merged). If the new Xinerama code
is present, then there is no need for the xinerama.conf server-side
writing/updating anymore.
this? I think it should be sufficient to make changes in x2goruncommand.
I think that too, see X2GoBTS #873 [1].
Should user still have possibility to turn Xinerama off? I can still
This is a good question. I don't think that we should include the
xinerama.conf part in nxagent anymore, but a way for switching off the
Xinerama extension should be possible, indeed.
The current plan is to make the cmdline options +xinerama / -xinerama
(or -extension Xinerama) functional in nxagent. With all previous
versions of nxagent, this did not have any effect, so far. Ulrich is
working on that.
imagine the use case when user have an x2goagent window in the middle of two physical monitors but want to see agent window as one display.
Indeed.
regards Alex
Mike
[1] https://bugs.x2go.org/cgi-bin/bugreport.cgi?bug=873
DAS-NETZWERKTEAM mike gabriel, herweg 7, 24357 fleckeby fon: +49 (1520) 1976 148
GnuPG Key ID 0x25771B31 mail: mike.gabriel@das-netzwerkteam.de, http://das-netzwerkteam.de
freeBusy: https://mail.das-netzwerkteam.de/freebusy/m.gabriel%40das-netzwerkteam.de.xf...
ok, let's wait till this feature is available in nxagent and I'll put support for it in x2go client.
regards, Alex
Am 18.05.2015 um 10:40 schrieb Mike Gabriel:
Hi Alex, (re-adding #872 to Cc:, dropping x2go-dev instead)
(my answers in-line at end of mail text)
On So 17 Mai 2015 08:45:57 CEST, Oleksandr Shneyder wrote:
Am 15.05.2015 um 13:59 schrieb Mike Gabriel:
Package: x2goclient Severity: important Version: 4.0.4.0
Hi Uli, hi Alex,
I turn this discussion into a bug report against X2Go Client.
Short summary for Alex: Ulrich has completely revisited [2] Xinerama in nx-libs recently for the 3.6.x branch of nx-libs [1].
The complete Xinerama code has been rewritten and moved mainly into the hw/nxagent/Screen.c file, so that we can soon drop your patched libNX_Xinerama.so library and all that LD_LIBRARY_PATH stuff from x2goruncommand.
With Ulrichs work, the hacky part in X2Go Client of writing and continuously updating a xinerama.conf file server-side from the client-side becomes obsolete and should only be kepts as legacy code for cases where X2Go Client talks to nxagent servers << 3.6.
I'd also appreciate it if you could give feedback on the below question.
Thanks, Mike
[1] https://github.com/ArcticaProject/nx-libs [2] https://github.com/ArcticaProject/nx-libs/pull/23
On Do 14 Mai 2015 23:32:03 CEST, Ulrich Sibiller wrote:
while testing my xinerama re-implementation I noticed that x2goclient reports a wrong width sometimes when simply moving the window without changing the size. The reported size differs by one pixel (too much or too little). When running x2goclient --debug you can see this:
x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 329, y: 214, w: 1440, h: 900" x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 365, y: 213, w: 1440, h: 900" x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 825, y: 190, w: 1440, h: 900" x2go-DEBUG-../src/sshprocess.cpp:189> executing remote command via SshProcess object (24): "export DISPLAY=:50;printf '\''1095 0 345 900\n0 0 1095 900'\'' > $HOME/.x2go/C-uli-50-1431636769_stDKDE_dp24/xinerama.conf"
-> here the 345 should have been a 344.
Interestingly my patched x2goagent (that completely ignores xinerama.conf) sees that value. So I conclude that x2goclient is manipulating the nxproxy window size by one pixel itself. Disabling xinerama in the session's configuration fixes the problem.
That off-by-one behaviour is caused by this code in onmainwindow.cpp:
void ONMainWindow::slotXineramaConfigured() { if (resumingSession.fullscreen) return; if (xinSizeInc == -1) xinSizeInc=1; else xinSizeInc=-1; #ifdef Q_OS_LINUX
lastDisplayGeometry.setWidth(lastDisplayGeometry.width()+xinSizeInc); XSync(QX11Info::display(),false); XResizeWindow(QX11Info::display(), proxyWinId,
lastDisplayGeometry.width(),lastDisplayGeometry.height()); XSync(QX11Info::display(),false); #endif #ifdef Q_OS_WIN QRect geom; wapiWindowRect ( (HWND) proxyWinId, geom ); wapiMoveWindow( (HWND) proxyWinId, geom.x(), geom.y(), geom.width()+xinSizeInc, geom.height(),true); lastDisplayGeometry=proxyWinGeometry(); #endif xineramaTimer->start(500); }
If I set xinSizeInc to zero in all cases the off-by-one behaviour is completely gone. Why does xinSizeInc exist at all?
It's a great news. Do we need to make a changes in the X2Go Client for
X2Go Client needs to test the remote x2goagent version and check whether it is << 3.5.99.0 or >= 3.5.99.0. If it is >= 3.5.99.0, the new Xinerama support will be present (once we have released 3.5.99.0, at the moment the code has not been merged). If the new Xinerama code is present, then there is no need for the xinerama.conf server-side writing/updating anymore.
this? I think it should be sufficient to make changes in x2goruncommand.
I think that too, see X2GoBTS #873 [1].
Should user still have possibility to turn Xinerama off? I can still
This is a good question. I don't think that we should include the xinerama.conf part in nxagent anymore, but a way for switching off the Xinerama extension should be possible, indeed.
The current plan is to make the cmdline options +xinerama / -xinerama (or -extension Xinerama) functional in nxagent. With all previous versions of nxagent, this did not have any effect, so far. Ulrich is working on that.
imagine the use case when user have an x2goagent window in the middle of two physical monitors but want to see agent window as one display.
Indeed.
regards Alex
Mike
--
Oleksandr Shneyder | Email: o.shneyder@phoca-gmbh.de phoca GmbH | Tel. : 0911 - 14870374 0 Ludwig-Feuerbach-str. 18 | Fax. : 0911 - 14870374 9 D-90489 Nürnberg | Mobil: 0163 - 49 64 461
Geschäftsführung: Dipl.-Inf. Oleksandr Shneyder
Amtsgericht München | http://www.phoca-gmbh.de HRB 196 658 | http://www.x2go.org USt-IdNr.: DE281977973
On Mon, May 18, 2015 at 11:23 AM, Oleksandr Shneyder <o.shneyder@phoca-gmbh.de> wrote:
ok, let's wait till this feature is available in nxagent and I'll put support for it in x2go client.
Ok. But can you please explain the reason for that off-by-one window size code I found (see below)?
On Do 14 Mai 2015 23:32:03 CEST, Ulrich Sibiller wrote:
while testing my xinerama re-implementation I noticed that x2goclient reports a wrong width sometimes when simply moving the window without changing the size. The reported size differs by one pixel (too much or too little). When running x2goclient --debug you can see this:
x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 329, y: 214, w: 1440, h: 900" x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 365, y: 213, w: 1440, h: 900" x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 825, y: 190, w: 1440, h: 900" x2go-DEBUG-../src/sshprocess.cpp:189> executing remote command via SshProcess object (24): "export DISPLAY=:50;printf '\''1095 0 345 900\n0 0 1095 900'\'' > $HOME/.x2go/C-uli-50-1431636769_stDKDE_dp24/xinerama.conf"
-> here the 345 should have been a 344.
Interestingly my patched x2goagent (that completely ignores xinerama.conf) sees that value. So I conclude that x2goclient is manipulating the nxproxy window size by one pixel itself. Disabling xinerama in the session's configuration fixes the problem.
That off-by-one behaviour is caused by this code in onmainwindow.cpp:
void ONMainWindow::slotXineramaConfigured() { if (resumingSession.fullscreen) return; if (xinSizeInc == -1) xinSizeInc=1; else xinSizeInc=-1; #ifdef Q_OS_LINUX
lastDisplayGeometry.setWidth(lastDisplayGeometry.width()+xinSizeInc); XSync(QX11Info::display(),false); XResizeWindow(QX11Info::display(), proxyWinId,
lastDisplayGeometry.width(),lastDisplayGeometry.height()); XSync(QX11Info::display(),false); #endif #ifdef Q_OS_WIN QRect geom; wapiWindowRect ( (HWND) proxyWinId, geom ); wapiMoveWindow( (HWND) proxyWinId, geom.x(), geom.y(), geom.width()+xinSizeInc, geom.height(),true); lastDisplayGeometry=proxyWinGeometry(); #endif xineramaTimer->start(500); }
If I set xinSizeInc to zero in all cases the off-by-one behaviour is completely gone. Why does xinSizeInc exist at all?
Uli
Hi Ulrich,
yes, if you moving window between two physical displays, without resizing it, x2goagent doesn't know, that display configuration is changed, because geometry of the x2goagent stays same. So I resized a window to refresh a display configurations.
regards, Alex Am 18.05.2015 um 13:28 schrieb Ulrich Sibiller:
On Mon, May 18, 2015 at 11:23 AM, Oleksandr Shneyder <o.shneyder@phoca-gmbh.de> wrote:
ok, let's wait till this feature is available in nxagent and I'll put support for it in x2go client.
Ok. But can you please explain the reason for that off-by-one window size code I found (see below)?
On Do 14 Mai 2015 23:32:03 CEST, Ulrich Sibiller wrote:
while testing my xinerama re-implementation I noticed that x2goclient reports a wrong width sometimes when simply moving the window without changing the size. The reported size differs by one pixel (too much or too little). When running x2goclient --debug you can see this:
x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 329, y: 214, w: 1440, h: 900" x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 365, y: 213, w: 1440, h: 900" x2go-DEBUG-../src/onmainwindow.cpp:10160> "New proxy geometry: (x: 825, y: 190, w: 1440, h: 900" x2go-DEBUG-../src/sshprocess.cpp:189> executing remote command via SshProcess object (24): "export DISPLAY=:50;printf '\''1095 0 345 900\n0 0 1095 900'\'' > $HOME/.x2go/C-uli-50-1431636769_stDKDE_dp24/xinerama.conf"
-> here the 345 should have been a 344.
Interestingly my patched x2goagent (that completely ignores xinerama.conf) sees that value. So I conclude that x2goclient is manipulating the nxproxy window size by one pixel itself. Disabling xinerama in the session's configuration fixes the problem.
That off-by-one behaviour is caused by this code in onmainwindow.cpp:
void ONMainWindow::slotXineramaConfigured() { if (resumingSession.fullscreen) return; if (xinSizeInc == -1) xinSizeInc=1; else xinSizeInc=-1; #ifdef Q_OS_LINUX
lastDisplayGeometry.setWidth(lastDisplayGeometry.width()+xinSizeInc); XSync(QX11Info::display(),false); XResizeWindow(QX11Info::display(), proxyWinId,
lastDisplayGeometry.width(),lastDisplayGeometry.height()); XSync(QX11Info::display(),false); #endif #ifdef Q_OS_WIN QRect geom; wapiWindowRect ( (HWND) proxyWinId, geom ); wapiMoveWindow( (HWND) proxyWinId, geom.x(), geom.y(), geom.width()+xinSizeInc, geom.height(),true); lastDisplayGeometry=proxyWinGeometry(); #endif xineramaTimer->start(500); }
If I set xinSizeInc to zero in all cases the off-by-one behaviour is completely gone. Why does xinSizeInc exist at all?
Uli
--
Oleksandr Shneyder | Email: o.shneyder@phoca-gmbh.de phoca GmbH | Tel. : 0911 - 14870374 0 Ludwig-Feuerbach-str. 18 | Fax. : 0911 - 14870374 9 D-90489 Nürnberg | Mobil: 0163 - 49 64 461
Geschäftsführung: Dipl.-Inf. Oleksandr Shneyder
Amtsgericht München | http://www.phoca-gmbh.de HRB 196 658 | http://www.x2go.org USt-IdNr.: DE281977973
On Mon, May 18, 2015 at 3:41 PM, Oleksandr Shneyder <o.shneyder@phoca-gmbh.de> wrote:
yes, if you moving window between two physical displays, without resizing it, x2goagent doesn't know, that display configuration is changed, because geometry of the x2goagent stays same. So I resized a window to refresh a display configurations.
Why does nxagent/x2goagent need know about that? In the current solution with xinerama.conf I don't see why x2goagent should be informed in this case. And if it really needs to know, this patch _should_ help (I have not tested this patch alone):
https://github.com/uli42/nx-libs/commit/61018cd39d02fbda1f2f766d8320c53866fa...
Uli
Imagine that you have agent window in the middle of two screens. Window size is 1024x768. You current displays configuration in that case is: left-512x768, right-512x768. Imagine, that you moved your window 100 px left. Now your displays configuration is: left-612x768, right-412x768. But from the view of x2goagent nothing is changed, agent has no idea about the display geometry on the client side. This is why client updating xinerama.conf and resizing a window to apply new settings. I hope it's understandable.
regards, Alex
Am 18.05.2015 um 21:01 schrieb Ulrich Sibiller:
On Mon, May 18, 2015 at 3:41 PM, Oleksandr Shneyder <o.shneyder@phoca-gmbh.de> wrote:
yes, if you moving window between two physical displays, without resizing it, x2goagent doesn't know, that display configuration is changed, because geometry of the x2goagent stays same. So I resized a window to refresh a display configurations.
Why does nxagent/x2goagent need know about that? In the current solution with xinerama.conf I don't see why x2goagent should be informed in this case. And if it really needs to know, this patch _should_ help (I have not tested this patch alone):
https://github.com/uli42/nx-libs/commit/61018cd39d02fbda1f2f766d8320c53866fa...
Uli
--
Oleksandr Shneyder | Email: o.shneyder@phoca-gmbh.de phoca GmbH | Tel. : 0911 - 14870374 0 Ludwig-Feuerbach-str. 18 | Fax. : 0911 - 14870374 9 D-90489 Nürnberg | Mobil: 0163 - 49 64 461
Geschäftsführung: Dipl.-Inf. Oleksandr Shneyder
Amtsgericht München | http://www.phoca-gmbh.de HRB 196 658 | http://www.x2go.org USt-IdNr.: DE281977973
On Tue, May 19, 2015 at 10:09 AM, Oleksandr Shneyder <o.shneyder@phoca-gmbh.de> wrote:
Imagine that you have agent window in the middle of two screens. Window size is 1024x768. You current displays configuration in that case is: left-512x768, right-512x768. Imagine, that you moved your window 100 px left. Now your displays configuration is: left-612x768, right-412x768. But from the view of x2goagent nothing is changed, agent has no idea about the display geometry on the client side. This is why client updating xinerama.conf and resizing a window to apply new settings. I hope it's understandable.
Yes, it is understandable. But this only relevant for xinerama information. And that information is coming from the config file (which is correctly updated in this case by x2goclient). So I still do not see why x2goagent needs to be triggered in any way. If a software needs to know the xinerama setup it issues a XineramaQueryScreens() request. This is intercepted by the libNX_xinerama.so and answered with the information from the config file. The x2goagent does not even see this request and thus does noot need to be updated. The only problem that your code solves might be the window manager in the session that caches xinerama information.
Do you have a use case that does not work without the resize?
Uli
Yes, It was exactly the problem, the different WM didn't update the configuration without resizing.
Am 19.05.2015 um 10:50 schrieb Ulrich Sibiller:
On Tue, May 19, 2015 at 10:09 AM, Oleksandr Shneyder <o.shneyder@phoca-gmbh.de> wrote:
Imagine that you have agent window in the middle of two screens. Window size is 1024x768. You current displays configuration in that case is: left-512x768, right-512x768. Imagine, that you moved your window 100 px left. Now your displays configuration is: left-612x768, right-412x768. But from the view of x2goagent nothing is changed, agent has no idea about the display geometry on the client side. This is why client updating xinerama.conf and resizing a window to apply new settings. I hope it's understandable.
Yes, it is understandable. But this only relevant for xinerama information. And that information is coming from the config file (which is correctly updated in this case by x2goclient). So I still do not see why x2goagent needs to be triggered in any way. If a software needs to know the xinerama setup it issues a XineramaQueryScreens() request. This is intercepted by the libNX_xinerama.so and answered with the information from the config file. The x2goagent does not even see this request and thus does noot need to be updated. The only problem that your code solves might be the window manager in the session that caches xinerama information.
Do you have a use case that does not work without the resize?
Uli
--
Oleksandr Shneyder | Email: o.shneyder@phoca-gmbh.de phoca GmbH | Tel. : 0911 - 14870374 0 Ludwig-Feuerbach-str. 18 | Fax. : 0911 - 14870374 9 D-90489 Nürnberg | Mobil: 0163 - 49 64 461
Geschäftsführung: Dipl.-Inf. Oleksandr Shneyder
Amtsgericht München | http://www.phoca-gmbh.de HRB 196 658 | http://www.x2go.org USt-IdNr.: DE281977973
On Tue, May 19, 2015 at 10:54 AM, Oleksandr Shneyder <o.shneyder@phoca-gmbh.de> wrote:
Yes, It was exactly the problem, the different WM didn't update the configuration without resizing.
This should be solved with the posted nxagent patch which will also be included in my xinerama rewrite. However, I'd suggest to modify the current x2goclient code to perform TWO resizes in order to achieve the same configuration as before. I had situations where the window stayed with on pixel too big or too small.
Uli
Hi Alex,
On Di 19 Mai 2015 10:09:54 CEST, Oleksandr Shneyder wrote:
Am 18.05.2015 um 21:01 schrieb Ulrich Sibiller:
On Mon, May 18, 2015 at 3:41 PM, Oleksandr Shneyder <o.shneyder@phoca-gmbh.de> wrote:
yes, if you moving window between two physical displays, without resizing it, x2goagent doesn't know, that display configuration is changed, because geometry of the x2goagent stays same. So I resized a window to refresh a display configurations.
Why does nxagent/x2goagent need know about that? In the current solution with xinerama.conf I don't see why x2goagent should be informed in this case. And if it really needs to know, this patch _should_ help (I have not tested this patch alone):
https://github.com/uli42/nx-libs/commit/61018cd39d02fbda1f2f766d8320c53866fa...
Uli
Imagine that you have agent window in the middle of two screens. Window size is 1024x768. You current displays configuration in that case is: left-512x768, right-512x768. Imagine, that you moved your window 100 px left. Now your displays configuration is: left-612x768, right-412x768. But from the view of x2goagent nothing is changed, agent has no idea about the display geometry on the client side. This is why client updating xinerama.conf and resizing a window to apply new settings. I hope it's understandable.
This is exactly what Ulrich is working on [1]. In the first test
builds with his patches, the nxagent window adapts the Xinerama layout
inside the nxagent whenever nxagent is moved over the client-side
screen area. Especially check the changes in hw/nxagent/Screen.c, and
there the intersect() function.
With Ulrich's patches, nxagent has Xinerama support without being in
need of a xinerama.conf file. I tested that with Python X2Go (i.e.,
via nxproxy) and via command line session startups (i.e., all on the
same host).
There are still some kinks we observe with several window managers,
but I am pretty sure that those issues need to be fixed in the window
managers (esp. MATE's window manager and also openbox).
Mike
https://github.com/ArcticaProject/nx-libs/pull/23/files
DAS-NETZWERKTEAM mike gabriel, herweg 7, 24357 fleckeby fon: +49 (1520) 1976 148
GnuPG Key ID 0x25771B31 mail: mike.gabriel@das-netzwerkteam.de, http://das-netzwerkteam.de
freeBusy: https://mail.das-netzwerkteam.de/freebusy/m.gabriel%40das-netzwerkteam.de.xf...
Now that Xinerama functionality has been reimplemented in nx-libs 3.6.x, we shall add an option to X2Go Client session configs to *disable* Xinerama.
XINERAMA is now enabled by default. We want to keep that default and not enable this option by default in X2Go Client. This is different from the old behavior, which defaulted to XINERAMA being off and thus had to be explicitly enabled.
Code that created and periodically updated the xinerama.conf file shall be kept for legacy versions of nx-libs, but is not necessary with nx-libs 3.6.x anymore.
See #873 for how XINERAMA can be disabled.
Mihai
Hi Mihai,
On Mi 13 Jul 2016 06:06:16 CEST, Mihai Moldovan wrote:
Code that created and periodically updated the xinerama.conf file
shall be kept for legacy versions of nx-libs, but is not necessary with nx-libs
3.6.x anymore.
The above can be done conditionally:
x2gofeature X2GOAGENT_RANDRXINERAMA
should return "ok" on stdout, if the feature is present.
In Python X2Go I load the "x2gofeaturelist" on every server
connection, so that I know what features are supported and what not.
Note, that X2Go Client so far has not been using this
x2gofeature(list) thingy. I wrote that for Python X2Go to be more
flexible on the client side when it comes to new server features and
legacy support for older servers.
Also note, that x2goserver-x2goagent for Debian was actually buggy
regarding that new feature, see:
http://code.x2go.org/gitweb?p=x2goserver.git;a=commitdiff;h=43510ec2ddd5d0fc...
Mike
--
DAS-NETZWERKTEAM mike gabriel, herweg 7, 24357 fleckeby mobile: +49 (1520) 1976 148 landline: +49 (4354) 8390 139
GnuPG Fingerprint: 9BFB AEE8 6C0A A5FF BF22 0782 9AF4 6B30 2577 1B31 mail: mike.gabriel@das-netzwerkteam.de, http://das-netzwerkteam.de
-
Mihai Moldovan -
Mike Gabriel -
Oleksandr Shneyder -
Ulrich Sibiller -
Ulrich Sibiller