X-Git-Url: http://git.annexia.org/?p=libguestfs.git;a=blobdiff_plain;f=BUGS;h=287586a7a0e439a325e8d7069688298f2431b9a8;hp=7acddb068e4021cbd55b8b4ecadd3f9c7402f211;hb=41b959fd9b057354f642d84398b875d02b88b864;hpb=ecee63de60a2419abf54eb1967dab04643234ed6

diff --git a/BUGS b/BUGS
index 7acddb0..287586a 100644
--- a/BUGS
+++ b/BUGS
@@ -1,8 +1,32 @@
 [guestfish] Built-in commands like 'alloc' and 'help' don't auto-complete.
 
 [libguestfs] /tmp/libguestfs temporary directory is not cleaned up
-(sometimes).
+sometimes.  This appears to happen (only?) if the main program
+segfaults.
+
+[libguestfs] qemu process not cleaned up occasionally.  This appears
+to happen only if the main program or library segfaults, in which case
+the atexit handler which would normally perform cleanup is not run.
+sometimes.  This appears to happen (only?) if the main program
+segfaults.
+[testing a hacky fix to this]
 
 [kernel?] Kernel boot times are significantly slower for Fedora 11/12
 than for Fedora 10, like 2-3 times slower.  This greatly affects the
 time it takes to launch the subprocess (although it does still work).
+[fixed - was a KVM bug]
+
+[java] Java bindings documentation messed up.  See:
+http://java.sun.com/j2se/javadoc/writingdoccomments/
+
+[libguestfs] If qemu process is a shell script, then it is not cleaned
+up correctly, even on "good" exit.  This is because we need to set up
+a new process group (see commented code for doing setpgid).
+[fixed? - added documentation in the manual page that you have to
+do 'exec qemu' in the wrapper.  However I still notice that the
+recovery process kills qemu, instead of it being killed by the
+parent process, or at least that's what I think is happening]
+
+[libguestfs] Some occasional data corruption, certainly on tgz-out.
+Try using tgz-out on a very large directory.
+[we think this is fixed by 64a6a828bd318622420cf3018899392fd80c14e7]