Bug 207301 - editors/openoffice-devel: Does not end on Ctrl-Q or file menu if Quickstarter in systray is enabled
Summary: editors/openoffice-devel: Does not end on Ctrl-Q or file menu if Quickstarter...
Status: Closed FIXED
Alias: None
Product: Ports & Packages
Classification: Unclassified
Component: Individual Port(s) (show other bugs)
Version: Latest
Hardware: Any Any
: --- Affects Only Me
Assignee: FreeBSD Office Team
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2016-02-18 10:11 UTC by Matthias Apitz
Modified: 2016-03-21 05:55 UTC (History)
2 users (show)

See Also:
bugzilla: maintainer-feedback? (office)


Attachments
patch-main_sal_osl_unx_pipe.c (1.65 KB, text/x-csrc)
2016-02-18 10:11 UTC, Matthias Apitz
no flags Details
patch-main_sal_osl_unx_sockimpl.h (367 bytes, patch)
2016-02-18 10:12 UTC, Matthias Apitz
no flags Details | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Matthias Apitz 2016-02-18 10:11:15 UTC
Created attachment 167143 [details]
patch-main_sal_osl_unx_pipe.c

I have compiled the port editors/openoffice-devel (which brings
openoffice-4.2.1712473) ... the socket.c change was already included in the
source, but not the pipe.c change; I elaborated the attached two patch
files for AOO and recompiled; but
/usr/local/openoffice-4.2.1712473/openoffice4/program/soffice.bin
hangs on Ctrl-Q in accept; a gdb bt of the hanging thread shows:

(gdb) bt
#0  0x00000008018a141a in _accept () from /lib/libc.so.7
#1  0x00000008015a6936 in __thr_accept (s=8, addr=0x0, addrlen=0x0)
    at /usr/local/r292778/src/lib/libthr/thread/thr_syscalls.c:106
#2  0x000000080085479e in osl_acceptPipe ()
   from /usr/local/openoffice-4.2.1712473/openoffice4/program/libuno_sal.so.3
#3  0x000000080567f894 in vos::OPipe::accept ()
   from /usr/local/openoffice-4.2.1712473/openoffice4/program/libvos3gcc3.so
#4  0x0000000800c5e4cb in ?? ()
   from /usr/local/openoffice-4.2.1712473/openoffice4/program/libsofficeapp.so
#5  0x000000080567badf in vos::threadWorkerFunction_impl ()
   from /usr/local/openoffice-4.2.1712473/openoffice4/program/libvos3gcc3.so
#6  0x000000080084ecb5 in osl_setThreadTextEncoding ()
   from /usr/local/openoffice-4.2.1712473/openoffice4/program/libuno_sal.so.3
#7  0x00000008015a4824 in thread_start (curthread=0x80cfee000)
    at /usr/local/r292778/src/lib/libthr/thread/thr_create.c:288
#8  0x0000000000000000 in ?? ()

i.e. the typical situation, as the pipe.c patch wasn't in effect; but it
is; I think osl_closePipe() is not called for some reason...

I went debugging through the frames (with printf) and can verify that
the Ctrl-Q is still visible in work/aoo-4.2.0/main/vcl/source/window/winproc.cxx
and we see on pressing 'Ctrl' the nEvent 0x0007 and on 'Q' the nEvent 0x005
(I grabbed the defines from work/aoo-4.2.0/main/vcl/inc/salwtype.hxx); I
pressed only 'Ctrl' two times and on the second I pressed 'Q' while
holding down 'Ctrl':

returned from Application::Yield() pSVData->maAppData.mbAppQuit: 0
calling Application::Yield() for the 127 time
ImplWindowFrameProc() nEvent: 0007
ImplWindowFrameProc() we have SALEVENT_KEYMODCHANGE and do ImplHandleSalKeyMod()
returned from Application::Yield() pSVData->maAppData.mbAppQuit: 0
calling Application::Yield() for the 128 time
ImplWindowFrameProc() nEvent: 0007
ImplWindowFrameProc() we have SALEVENT_KEYMODCHANGE and do ImplHandleSalKeyMod()
returned from Application::Yield() pSVData->maAppData.mbAppQuit: 0
calling Application::Yield() for the 129 time
returned from Application::Yield() pSVData->maAppData.mbAppQuit: 0
calling Application::Yield() for the 130 time
ImplWindowFrameProc() nEvent: 0005
ImplWindowFrameProc() we have SALEVENT_KEYINPUT and do ImplHandleKey()
returned from Application::Yield() pSVData->maAppData.mbAppQuit: 0
calling Application::Yield() for the 131 time
ImplWindowFrameProc() nEvent: 0016
Comment 1 Matthias Apitz 2016-02-18 10:12:13 UTC
Created attachment 167144 [details]
patch-main_sal_osl_unx_sockimpl.h
Comment 2 Don Lewis freebsd_committer freebsd_triage 2016-02-18 23:13:25 UTC
Have you disabled the Quickstarter?  I see that behavior of the Quickstarter is enabled.  If I start OpenOffice from the command line, the OpenOffice window disappears when I quit, but I don't get a command prompt because the process is still running as the Quickstarter, which I see as an icon in the toolbar.  If I click on the icon and either exit or disable the Quickstarter, the process exits and I get the command prompt.  If the Quickstarter is disabled, I get the command prompt back immediately when quitting OpenOffice.

Quickstarter can be enabled or disabled by navigating to Tools->Options->OpenOffice->Memory->Enable systray Quickstarter.

I think the LibreOffice problem was different because it is built by default with the QuickStarter disabled.
Comment 3 Matthias Apitz 2016-02-19 06:16:27 UTC
Wow! Yes, this solved the problem. Thank you. If you would know, how many hours I spent in debugging this, and I do not even use AOO, only LO. I only wanted to nail it down because I thought it was related to the problem I faced in LO...

Is this "feature" (I would call it rather a bug) somewhere documented as "If you enable systray, a started AOO on command line will never end"?

Never the less, the attached patches should be added to the port editors/openoffice-devel (...) as long as the pipe.c bug is not fixed upstream.
Comment 4 Don Lewis freebsd_committer freebsd_triage 2016-03-20 18:06:27 UTC
I'm in the process of updating openoffice-devel to upstream r1735544 which contains this change, but it still does not exit if I enable Quickstarter.  It appears that the Quickstarter does no daemonize itself.

If I start OpenOffice, enable Quickstarter, quit, kill the process, log out, and log in, then things work properly.  Enabling Quickstarter also flags OpenOffice as a startup application, so a copy is started in the background and I see the toolbar icon pop up when I log in.  If I then start OpenOffice from the command line I get my command prompt back right away.  That is the way that OpenOffice has behaved when opening a second document from the command line.
Comment 5 commit-hook freebsd_committer freebsd_triage 2016-03-20 21:43:07 UTC
A commit references this bug:

Author: truckman
Date: Sun Mar 20 21:42:08 UTC 2016
New revision: 411522
URL: https://svnweb.freebsd.org/changeset/ports/411522

Log:
  Delay running gnome-post-install until the plist has been generated by
  post-install so that the icons are detected and gtk-update-icon-cache
  gets run when the package is installed. [1]

  Import upstream trunk commit r1728872 to add the
  CLOSESOCKET_DOESNT_WAKE_UP_ACCEPT up accept fix to the pipe code. [2]

  PR:		208026 [1]
  PR:		207301 [2]

Changes:
  head/editors/openoffice-4/Makefile
  head/editors/openoffice-4/files/patch-r1728872
Comment 6 commit-hook freebsd_committer freebsd_triage 2016-03-21 05:53:52 UTC
A commit references this bug:

Author: truckman
Date: Mon Mar 21 05:53:01 UTC 2016
New revision: 411540
URL: https://svnweb.freebsd.org/changeset/ports/411540

Log:
  Upgrade to upstream svn revision r1735889.  This contains the
  CLOSESOCKET_DOESNT_WAKE_UP_ACCEPT pipe fix [1].  It also contains
  a slightly different fix for building with boost 1.60, so delete
  patch-sdext_source_pdfimport_makefile.mk, which is no longer needed.

  Delay running gnome-post-install until the plist has been generated by
  post-install so that the icons are detected and gtk-update-icon-cache
  gets run when the package is installed. [2]

  Add missing USES=shared-mime-info.

  PR:		207301 [1]
  PR:		208026 [2]

Changes:
  head/editors/openoffice-devel/Makefile
  head/editors/openoffice-devel/distinfo
  head/editors/openoffice-devel/files/patch-sdext_source_pdfimport_makefile.mk