Opened 14 years ago
Closed 13 years ago
#665 closed defect (fixed)
TGeom regress crash on Linux 64 bits
Reported by: | colivier | Owned by: | coliver |
---|---|---|---|
Priority: | blocker | Milestone: | PostGIS 2.0.0 |
Component: | postgis | Version: | master |
Keywords: | TGeom Tin PolyhedralSurface | Cc: |
Description
Cunit tin_tgeom and psurface_tgeom crash on 64 bits Linux arch. (theses 2 tests are right now deactivated)
Nota: this one seems not related to #664 as i can reproduce it with an old repo.
Change History (9)
comment:1 by , 13 years ago
Owner: | changed from | to
---|---|
Status: | new → assigned |
comment:2 by , 13 years ago
comment:4 by , 13 years ago
@coliver: how am I supposed to release memory associated with a TSERIALIZED ?
The memory errors occur on tgeom_deserialize inside check_tgeom.
comment:5 by , 13 years ago
The serializer isn't really serializing things, `data' remains hold by pointer
comment:6 by , 13 years ago
TSERIALIZED struct references tserialized.txt but I can't find it anywhere.
comment:7 by , 13 years ago
Owner: | changed from | to
---|---|
Priority: | medium → blocker |
Status: | assigned → new |
With r8337 I've renamed the suite and restricted the disabled code so it's clearer where the problems occur. I'll re-assign this to coliver as I guess he's the author of the code so he knows what's the serialized format supposed to be for surfaces.
Tagging this as a blocker while assuming the memory corrupting code path is exposed to SQL.
comment:8 by , 13 years ago
@coliver: are you planning to work on this any soon ? I'm basically missing documentation about how TSERIALIZED should look like so I am pretty blind.
Do you guys mind if I change the unit test suite names ? I've started already with some of them so to match filename. It helps when you're focusing on a single file as you can specify the name as an argument to cu_tester, as in:
Which is much easiear to type than: