| Commit message (Collapse) | Author | Age |
|
|
|
|
|
| |
Closes:#850079.
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This approach leaks the blocking signal mask into subprocesses,
It could be fixed by finding all the places where we fork and exec,
but vtwm likes to use system and popen, which are hard to adjust
since there is no hook which runs in the child. (We could use
pthread_atfork but then we'd have to link against pthreads...)
This reverts commit 91b2ee084430dbe5d81d32d15e06b53d931514d9.
(apart from the changelog entry)
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Previously we would interrupt whatever we were doing, willy-nilly.
Now we wait until the event loop is reentered. Symptoms could include
locking up the server with a grab, and/or spinning on the cpu.
This patch is not very portable (it uses ppoll) but I think it will
work on all Debian architectures.
In the longer term we should update to new upstream, where the last
commits were in 2013.
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
|
|
[dgit import orig vtwm_5.4.7.orig.tar.gz]
|