"libhal_device_free_changeset" function mem-leeks.
|Assignee:||Milan Jurik||% Done:|
|Category:||lib - userland libraries|
"LibHalChangeSetElement" struct contains "key" member, which is not freed by the function.
diff -r 48985d630b76 usr/src/lib/hal/libhal/common/libhal.c
--- a/usr/src/lib/hal/libhal/common/libhal.c Mon Jan 30 18:55:11 2012 +0100
+++ b/usr/src/lib/hal/libhal/common/libhal.c Fri Feb 03 08:57:57 2012 -0800
@@ -4182,6 +4182,7 @@
fprintf (stderr, "%s %d : unknown change_type %d\n", __FILE__, __LINE__, elem->change_type);
+ free (elem->key);
- Status changed from New to In Progress
- Assignee set to Milan Jurik
- Tags deleted (
It seems there are few leaks more in the code, I will look at them.
Why HAL is maintained in illumos-gate? It should be in userland.
Because all the important parts had to be written fresh for the Solaris port, and it (in theory, I haven't checked recently) consumes some deeply private interfaces.
- % Done changed from 0 to 20
Not only that. Upstream abandoned HAL, so all its development is on us now anyway. Some of Solaris specific parts were pushed to upstream but it does not seem all was done before HAL upstream stalled.
- % Done changed from 20 to 100
- Difficulty changed from Medium to Bite-size
Fix for RTI matches what Roman Strashkin proposed, it is from upstream.
The rest of possible leaks is covered by other issues.
- Status changed from In Progress to Pending RTI
- Status changed from Pending RTI to Resolved
Resolved in r13658 commit:1e6115622470
Also available in: Atom