Ticket #1015 (closed defect: fixed)

Opened 2 years ago

Last modified 2 years ago

[raster] segfault on make check

Reported by: strk Owned by: dustymugs
Priority: critical Milestone: PostGIS 2.0.0
Component: raster Version: trunk
Keywords: Cc:

Description

as of r7341, on a 64bit with gdal 1.9.0:

...
WARNING: Value set for 32BF band got converted from 98.989848 to 98.989845
WARNING: Value set for 32BF band got converted from 99.494949 to 99.494949
WARNING: Value set for 32BF band got converted from 99.994975 to 99.994972
ERROR 5: GDALWarpOptions.Validate()
  padfSrcNoDataReal set, but padfSrcNoDataImag not set.
make[3]: *** [check] Segmentation fault
make[3]: Leaving directory `/home/src/postgis/postgis/raster/test/core'

Attachments

bug1015.valgrind.txt Download (35.6 KB) - added by strk 2 years ago.

Change History

Changed 2 years ago by robe

This look like it's failing on the transform too -- ah we have a Linux victim suffering the same pains as a windows person :).

I'll double check on my end - but I think this is a dupe #925.

On windows under mingw, I believe I get the same error followed by can't load libproj-0.dll.

Strk -- to confirm -- can you look at that ticket and try the regress for transform on your linux box and confirm you are getting the wrong answers?

Changed 2 years ago by pracine

I get the exact same message on Windows.

Changed 2 years ago by robe

  • status changed from new to closed
  • resolution set to duplicate

okay I have marked this as dupe of 925

Changed 2 years ago by dustymugs

Which revision of GDAL trunk are you using? I'm using r22520.

Changed 2 years ago by dustymugs

On r7363 on a 64-bit linux box, I can't replicate that error with testapi.

Changed 2 years ago by strk

  • status changed from closed to reopened
  • resolution duplicate deleted

Still fails here as of r7366. GDAL is 22149.

Changed 2 years ago by dustymugs

Are you getting the same error message "padfSrcNoDataReal set, but padfSrcNoDataImag not set."?

Changed 2 years ago by strk

Nope, indeed that's gone. I'm just left with the segfault. Valgrind report attached (seem to reveal a few errors).

Changed 2 years ago by strk

Changed 2 years ago by dustymugs

Thanks for the valgrind. That is confirming what I'm seeing. I think I'm prematurely attempting to free memory.

Can you try r7367?

Changed 2 years ago by dustymugs

  • owner changed from pracine to dustymugs
  • status changed from reopened to new

Your posting a valgrind report got me moving to fix my glaring memory leaks in rt_api.c and testapi.c. I'll get along with valgrinding the full regression suite next.

Changed 2 years ago by dustymugs

strk, can you see if you still experience the seg fault? If you don't, please close this ticket. Thanks!

Changed 2 years ago by strk

Still fails as of r7397

*** glibc detected *** ./testapi: double free or corruption (out): 0x0000000000ca2550 ***
======= Backtrace: =========
/lib/libc.so.6(+0x775b6)[0x2b3bc92ed5b6]
/lib/libc.so.6(cfree+0x73)[0x2b3bc92f3e53]
/usr/local/lib/libgdal.so.1(VSIFree+0x1f)[0x2b3bc8b05038]
/usr/local/lib/libgdal.so.1(GDALDestroyWarpOptions+0x97)[0x2b3bc8b25c4d]
./testapi[0x4143b9]
./testapi[0x4084f6]
./testapi[0x40bf78]
/lib/libc.so.6(__libc_start_main+0xfd)[0x2b3bc9294c4d]
./testapi[0x4045b9]

Changed 2 years ago by dustymugs

Can you try again with r7399? It seems that someone else has had similar issues according to the discussion:

 http://lists.osgeo.org/pipermail/gdal-dev/2011-May/028730.html

Changed 2 years ago by strk

  • status changed from new to closed
  • resolution set to fixed

r7399 works here.

Note: See TracTickets for help on using tickets.