#1104 closed defect (invalid)
lwgeom_homogenize does a shallow copy
Reported by: | strk | Owned by: | pramsey |
---|---|---|---|
Priority: | medium | Milestone: | PostGIS 2.0.0 |
Component: | postgis | Version: | master |
Keywords: | Cc: |
Description
Is it intentional for lwgeom_homogenize to avoid a deep copy of coordinates ? Do we have a way to flag ownership of a collection components ?
Currently, the homogenize unit test fails with a double free of corruption as it lwgeom_free the input and the output passed to lwcollection_homogenize. How should memory be handled instead ?
Change History (2)
comment:1 by , 13 years ago
Resolution: | → invalid |
---|---|
Status: | new → closed |
comment:2 by , 13 years ago
For the record: r7579 is the commit which rewinded the overentusiastic memory release
Note:
See TracTickets
for help on using tickets.
Ok, I checked, lwcollection_homogenize uses lwgeom_clone, which in turn doesn't copy POINTARRAY memory (lwgeom_clone_deep would do that).
Then I guess this can't be called a bug on itself, but a general need to cleanup the memory management framework. For now I fixed the test to use lwgeom_release which supposedly does a shallow free (but still fails to drop some memory…)