test_rect_tree_contains_point segfaults

[strk]

Running test 'test_rect_tree_contains_point' in suite 'PostGIS Measures Suite'.

Program received signal SIGSEGV, Segmentation fault. 0x000000000040a217 in test_rect_tree_contains_point () at cu_measures.c:130 130 tree = rect_tree_new(poly→rings[0]); (gdb) bt #0 0x000000000040a217 in test_rect_tree_contains_point () at cu_measures.c:130 #1 0x00007f0a5821945a in ?? () from /usr/lib/libcunit.so.1 #2 0x00007f0a58219877 in CU_run_test () from /usr/lib/libcunit.so.1 #3 0x000000000040e2e4 in main (argc=2, argv≤value optimized out>)

at cu_tester.c:118

At revision 5443. Did ./autogen.sh; ./configure && make clean && make && make check

[strk]

Valgrind log:

==32102== Use of uninitialised value of size 8
==32102==    at 0x40A217: test_rect_tree_contains_point (cu_measures.c:130)
==32102==    by 0x50B3459: (within /usr/lib/libcunit.so.1.0.1)
==32102==    by 0x50B3876: CU_run_test (in /usr/lib/libcunit.so.1.0.1)
==32102==    by 0x40E2E3: main (cu_tester.c:118)
==32102== 
==32102== Invalid read of size 8
==32102==    at 0x40A217: test_rect_tree_contains_point (cu_measures.c:130)
==32102==    by 0x50B3459: (within /usr/lib/libcunit.so.1.0.1)
==32102==    by 0x50B3876: CU_run_test (in /usr/lib/libcunit.so.1.0.1)
==32102==    by 0x40E2E3: main (cu_tester.c:118)
==32102==  Address 0x0 is not stack'd, malloc'd or (recently) free'd
==32102== 

[pramsey]

There's not much code there… it's like the actual from EWKT process is failing? But that implies parser stuff is failing, which implies…?

[strk]

Replying to pramsey:

There's not much code there… it's like the actual from EWKT process is failing? But that implies parser stuff is failing, which implies…?

No idea Maybe that 'make clean' failed to clean some of the parser stuff ?

Versions here:

bison (GNU Bison) 2.3 flex 2.5.35

[pramsey]

I couldn't replicate this on 32-bit or 64-bit Centos linux.

[pramsey]

I couldn't replicate it by regenerating the parser files either. And in any event the generated files haven't been touched since November of last year.

[strk]

My segfault is also a moving target. This time it happened here:

Test: test_ptarray_point_in_ring … make[1]: * [check] Segmentation fault

[strk]

Alright, as of r5445 'make clean' does what it's supposed to do, and next build&check after make clean now succeeds with no failures nor segfaults.

I'm afraid dependencies are still a big mess, but at least you can clean now.