[SCM] Qt 4 Debian packaging branch, master, updated. debian/4.6.2-5-5-gf774321
Modestas Vainius
modax at alioth.debian.org
Sun Jun 6 01:02:35 UTC 2010
The following commit has been merged in the master branch:
commit f7743216a1cb91373bd37f05db3c12da992cf09e
Author: Modestas Vainius <modestas at vainius.eu>
Date: Sun Jun 6 03:36:37 2010 +0300
Decrease failure rate in QProcess on hppa caused by #561203.
Add usleep(1000) before fork() in QProcess on hppa in order to decrease failure
rate caused by #561203. Hopefully, probablity of the failure becomes so low
that packages do not FBTFS anymore. This hack is just an ugly but very needed
workaround for a common manifestation of the bug in QProcess (patch
99_hppa_bug561203_decrease_failure_rate.diff).
---
debian/changelog | 5 ++++
.../99_hppa_bug561203_decrease_failure_rate.diff | 24 ++++++++++++++++++++
2 files changed, 29 insertions(+), 0 deletions(-)
diff --git a/debian/changelog b/debian/changelog
index 8e22d37..a311d72 100644
--- a/debian/changelog
+++ b/debian/changelog
@@ -3,6 +3,11 @@ qt4-x11 (4:4.6.2-6) UNRELEASED; urgency=low
* Expand 96_webkit_no_gc_sections.diff to cover mips(el) as well. Hopefully,
this will make linker less memory demanding and solve out-of-memory
condition on buildds.
+ * Add usleep(1000) before fork() in QProcess on hppa in order to decrease
+ failure rate caused by #561203. Hopefully, probablity of the failure
+ becomes so low that packages do not FBTFS anymore. This hack is just an
+ ugly but very needed workaround for a common manifestation of the bug in
+ QProcess (patch 99_hppa_bug561203_decrease_failure_rate.diff).
-- Modestas Vainius <modax at debian.org> Sun, 06 Jun 2010 03:16:11 +0300
diff --git a/debian/patches/99_hppa_bug561203_decrease_failure_rate.diff b/debian/patches/99_hppa_bug561203_decrease_failure_rate.diff
new file mode 100644
index 0000000..8156f56
--- /dev/null
+++ b/debian/patches/99_hppa_bug561203_decrease_failure_rate.diff
@@ -0,0 +1,24 @@
+Description: decrease failure rate of QProcess (#561203) on hppa ~100+ times
+ The patch adds a usleep() of 1000 before fork() in QProcess on hppa. This hack
+ seems to bring down likelihood of failure caused by #561203 from ~ 1/6 to
+~1/1000 when run in a row. Hopefully, this will make failures rare enough
+ not to cause FTBFSes.
+ Degradation in performance should not be very significant and is acceptable.
+Author: Modestas Vainius <modax at debian.org>
+Forwarded: not-needed
+Bug-Debian: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=561203
+Origin: vendor
+Last-Update: 2010-06-06
+
+--- a/src/corelib/io/qprocess_unix.cpp
++++ b/src/corelib/io/qprocess_unix.cpp
+@@ -505,6 +505,9 @@ static char **_q_dupEnvironment(const QH
+ // under QNX RTOS we have to use vfork() when multithreading
+ inline pid_t qt_fork()
+ {
++#if defined(__linux__) && defined(__hppa__)
++ usleep(1000);
++#endif
+ #if defined(Q_OS_QNX)
+ return vfork();
+ #else
--
Qt 4 Debian packaging
More information about the pkg-kde-commits
mailing list