tillea
commented
5 years ago Hi, I have tracked down the issue in a gdb session:
ffindex_build -s ./test.data ./test.ffindex test/data test/data2
ffindex_get ./test.data ./test.ffindex a b foo | tee ./test.out
a
bb
fooo
fooo
gdb --args ffindex_modify -u ./test.ffindex b
GNU gdb (Debian 8.2-1) 8.2
...
(gdb) break src/ffindex.c:553
Breakpoint 1 (src/ffindex.c:553) pending.
(gdb) run
Starting program: ffindex_modify -u ./test.ffindex b
Breakpoint 1, action (node=0x555555559801, which=postorder, depth=1, misc=0x555555559260) at /build/ffindex-0.9.9.7.soedinglab+git20171201.74550c8/src/ffindex.c:553
553 entry = *(ffindex_entry_t **) node;
(gdb) print entry
$1 = (ffindex_entry_t *) 0x7
(gdb) print entry->offset
Cannot access memory at address 0x7
(gdb) next
554 if(fprintf(index_file, "%s\t%zd\t%zd\n", entry->name, entry->offset, entry->length) < 0)
(gdb) next
Program received signal SIGSEGV, Segmentation fault.
0x00007ffff7f7bc6d in action (node=0x555555559801, which=postorder, depth=1, misc=0x555555559260) at /build/ffindex-0.9.9.7.soedinglab+git20171201.74550c8/src/ffindex.c:554
554 if(fprintf(index_file, "%s\t%zd\t%zd\n", entry->name, entry->offset, entry->length) < 0)So the SIGSEGV occures in the attempt to access any member of the struct entry (I have some more detailed debugging code which proves this statement). Another thing to consider is that this SIGSEGV occures only for gcc 8.2. When using Debian stable featuring gcc 6.3 the problem does not occure and the test runs fine. Hope this helps.
ahcm
Hi, I just learned in Issue 4 that I might have grabbed the wrong codebase for Debian when I pointed it to this Git repository. However, you might be interested to know that ffindex_modify segfaults in our test suite which you can see in the Debian bug report. Hope this helps, Andreas.