Integrate with Valgrind

[smtakeda]

No description provided.

[smtakeda]

PHP tests can run with valgrind by

TEST_PHP_ARGS=-m make test

But the large result set tests will hit a test timeout as it takes more than 5 min.

[smtakeda]

ctests raised this. This doesn't happen in the local environment. Not sure what is going on.

==9738== Invalid free() / delete / delete[] / realloc()
==9738==    at 0x4C2BDEC: free (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==9738==    by 0x57E2DEB: __libc_freeres (in /lib/x86_64-linux-gnu/libc-2.19.so)
==9738==    by 0x4A256BC: _vgnU_freeres (in /usr/lib/valgrind/vgpreload_core-amd64-linux.so)
==9738==    by 0x56B616A: __run_exit_handlers (exit.c:97)
==9738==    by 0x56B61F4: exit (exit.c:104)
==9738==    by 0x569BF4B: (below main) (libc-start.c:321)
==9738==  Address 0x5a3c3c0 is 0 bytes inside data symbol "noai6ai_cached"
==9738== 
==9738== 
==9738== HEAP SUMMARY:
==9738==     in use at exit: 0 bytes in 0 blocks
==9738==   total heap usage: 84,111 allocs, 84,112 frees, 8,407,676 bytes allocated
==9738== 
==9738== All heap blocks were freed -- no leaks are possible
==9738== 
==9738== For counts of detected and suppressed errors, rerun with: -v
==9738== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0)

[smtakeda]

--run-libc-freeres=no removed the error. Looks like it was due to old glibc.

[smtakeda]

All tests but one test passed.
https://travis-ci.org/snowflakedb/pdo_snowflake/jobs/325963856
This is the large result set test. @notkriswagner do you have any idea about this?

==10080== 
==10080== HEAP SUMMARY:
==10080==     in use at exit: 8,285 bytes in 58 blocks
==10080==   total heap usage: 889,779 allocs, 889,721 frees, 94,992,156,847 bytes allocated
==10080== 
==10080== 544 bytes in 2 blocks are possibly lost in loss record 20 of 25
==10080==    at 0x4C2CC70: calloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==10080==    by 0x4012FE4: allocate_dtv (dl-tls.c:296)
==10080==    by 0x4012FE4: _dl_allocate_tls (dl-tls.c:460)
==10080==    by 0x5464D92: allocate_stack (allocatestack.c:589)
==10080==    by 0x5464D92: pthread_create@@GLIBC_2.2.5 (pthread_create.c:500)
==10080==    by 0x45C523: chunk_downloader_init (chunk_downloader.c:285)
==10080==    by 0x45188B: snowflake_execute (client.c:1447)
==10080==    by 0x44FF5F: snowflake_query (client.c:939)
==10080==    by 0x44D433: main (large_result_set.c:31)
==10080== 
==10080== LEAK SUMMARY:
==10080==    definitely lost: 0 bytes in 0 blocks
==10080==    indirectly lost: 0 bytes in 0 blocks
==10080==      possibly lost: 544 bytes in 2 blocks
==10080==    still reachable: 7,741 bytes in 56 blocks
==10080==         suppressed: 0 bytes in 0 blocks
==10080== Reachable blocks (those to which a pointer was found) are not shown.
==10080== To see them, rerun with: --leak-check=full --show-leak-kinds=all
==10080== 
==10080== For counts of detected and suppressed errors, rerun with: -v
==10080== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0)

Will look into this anyway.

[smtakeda]

If I ran the large result set test, no valgrind error is raised:

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
Number of rows in result: 100000
Number of rows fetched: 100000
Number of rows fetched equals number of rows expected in result
==27963== 
==27963== HEAP SUMMARY:
==27963==     in use at exit: 0 bytes in 0 blocks
==27963==   total heap usage: 759,688 allocs, 759,688 frees, 101,659,792,885 bytes allocated
==27963== 
==27963== All heap blocks were freed -- no leaks are possible
==27963== 
==27963== For counts of detected and suppressed errors, rerun with: -v
==27963== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 8 from 6)

[smtakeda]

Now it is reproducible on the local vm. Moving the ticket to libsnowflakeclient.

[smtakeda]

--run-libc-freeres=no caused the above issue. That's false alarm!