Apache OpenOffice (AOO) Bugzilla – Issue 63397
SDK: examples does not work on Solaris
Last modified: 2016-07-19 17:49:19 UTC
While compiling examples/DevelopersGuide/Components/CppComponent the unopkg crashes on Solaris.
.
ok in sb49
See issue 63473 for details.
changes made: bridges/source/remote/context/context.cxx:1.6.20.1 cppu/prj/build.lst:1.10.40.1 cppu/source/threadpool/jobqueue.cxx:1.3.38.1 cppu/source/threadpool/jobqueue.hxx:1.3.38.1 cppu/source/threadpool/thread.cxx:1.7.38.1 cppu/source/threadpool/thread.hxx:1.4.40.1 cppu/source/threadpool/threadpool.cxx:1.12.38.1 cppu/source/threadpool/threadpool.hxx:1.2.38.1 cppu/source/uno/lbmap.cxx:1.23.38.1 cppu/util/makefile.mk:1.7.40.1
reopen
back to sb
JSC -> SB: Stephan i reopened the bug because i moved your changes back because of a UNO remote problem (second connect failed, C++ and Java and probaly other languges as well). Internal P1 issue 134738 -> automatic testtool doesn't work i have changed back: cppu/source/threadpool/jobqueue.cxx:1.4 to1.3 cppu/source/threadpool/jobqueue.hxx:1.4 to 1.3 cppu/source/threadpool/thread.cxx:1.8 to 1.7 cppu/source/threadpool/thread.hxx:1.5 to 1.4 cppu/source/threadpool/threadpool.cxx:1.13 to 1.12 cppu/source/threadpool/threadpool.hxx:1.3 to 1.2
=> 2.0.4
cppu/source/threadpool/threadpool.cxx:1.13 fails to take into account that uno_threadpool_destroy might be followed by uno_threadpool_create again (if one remote connection is used and terminated, and then another remote connection is used). A simple g_threadPool=0 at the end of uno_threadpool_destroy is not enough, however: For one, g_threadPool must be set to zero only if its refcount reaches zero, and for another, uno_threadpool_attach/detach probably also need to adjust g_threadPool's refcount to synchronize with another thread calling uno_threadpool_destroy.
target 2.x
Issue 80121 is probably a duplicate of this issue. Remember to undo that issue's "temporary fix [...] to dispose the bridges directly."
see also issue 111223
The changes made to bridges/source/remote/context/context.cxx:1.6.20.1 have effectively been reverted through issue 112784.
Confusing comments but it seems this issue was either resolved or a duplicate of another.