[
https://issues.apache.org/jira/browse/ARROW-7553?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Mike Pontillo updated ARROW-7553:
---------------------------------
Component/s: Documentation
Priority: Minor (was: Major)
> Unable to free all memory allocated when creating a GArrowSchema
> ----------------------------------------------------------------
>
> Key: ARROW-7553
> URL: https://issues.apache.org/jira/browse/ARROW-7553
> Project: Apache Arrow
> Issue Type: Bug
> Components: C, Documentation, GLib
> Affects Versions: 0.15.1
> Environment: Ubuntu Focal with Linuxbrew packages installed
> Reporter: Mike Pontillo
> Priority: Minor
>
> I was not able to run the test code in [this
> gist|https://gist.github.com/mpontillo/cfbe7ebbf0b0f2acf31063512439bde7] (and
> corresponding [GitHub issue|https://github.com/apache/arrow/issues/6164])
> without encountering a memory leak via {{garrow_schema_get_type_once}}
> (according to {{valgrind}}):
> {code:java}
> 24 bytes in 1 blocks are possibly lost in loss record 192 of 486
> at 0x4A391AF: realloc (vg_replace_malloc.c:836)
> by 0x53B5D97: g_realloc (gmem.c:164)
> by 0x5128108: type_node_any_new_W (gtype.c:502)
> by 0x512D9EC: g_type_register_static (gtype.c:2766)
> by 0x512DDB4: g_type_register_static_simple (gtype.c:2719)
> by 0x4CB3CF8: garrow_schema_get_type_once (in
> /home/linuxbrew/.linuxbrew/Cellar/apache-arrow-glib/0.15.1_1/lib/libarrow-glib.so.15.0.1)
> by 0x4CB3E40: garrow_schema_get_type (in
> /home/linuxbrew/.linuxbrew/Cellar/apache-arrow-glib/0.15.1_1/lib/libarrow-glib.so.15.0.1)
> by 0x4CB4098: garrow_schema_new_raw(std::shared_ptr<arrow::Schema>*) (in
> /home/linuxbrew/.linuxbrew/Cellar/apache-arrow-glib/0.15.1_1/lib/libarrow-glib.so.15.0.1)
> by 0x4CB5860: garrow_schema_new (in
> /home/linuxbrew/.linuxbrew/Cellar/apache-arrow-glib/0.15.1_1/lib/libarrow-glib.so.15.0.1)
> by 0x401219: main (main.c:13)
> {code}
> I'm not positive if this is an actual bug in {{apache-arrow-glib}} or if my
> test code is incorrect, but I thought I would file an issue here (in addition
> to my question posted as a GitHub issue) for greater visibility.
> It looks like the {{realloc}} call actually happens via {{glib}}; is this a
> one-time type registration that isn't expected to be cleaned up until the
> process exits? (I couldn't find a reference to
> {{garrow_schema_get_type_once}} in the {{arrow}} code.) I wanted to be
> certain, since I expect to use {{arrow-glib}} within a long-running process.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
