| Bug #62572 | "make test" fails on "my_vsnprintf" | ||
|---|---|---|---|
| Submitted: | 29 Sep 2011 11:19 | Modified: | 3 Jun 2013 14:09 |
| Reporter: | Oleksii Moskvin | Email Updates: | |
| Status: | Closed | Impact on me: | |
| Category: | Tests | Severity: | S3 (Non-critical) |
| Version: | 5.5.16, 5.5.20, 5.5.21 | OS: | Linux |
| Assigned to: | CPU Architecture: | Any | |
| Tags: | make test, my_vsnprintf, unit tests | ||
[29 Sep 2011 12:54]
MySQL Verification Team
I couldn't repeat with source tree code 1 week older:
miguel@lara ~/mysql-5.5 $ make test
Running tests...
Test project /home/miguel/mysql-5.5
Start 1: pfs_instr_class
1/12 Test #1: pfs_instr_class .................. Passed 0.05 sec
Start 2: pfs_instr_class-oom
2/12 Test #2: pfs_instr_class-oom .............. Passed 0.06 sec
Start 3: pfs_instr
3/12 Test #3: pfs_instr ........................ Passed 0.05 sec
Start 4: pfs_instr-oom
4/12 Test #4: pfs_instr-oom .................... Passed 0.05 sec
Start 5: pfs
5/12 Test #5: pfs .............................. Passed 0.63 sec
Start 6: bitmap
6/12 Test #6: bitmap ........................... Passed 7.14 sec
Start 7: base64
7/12 Test #7: base64 ........................... Passed 0.12 sec
Start 8: my_vsnprintf
8/12 Test #8: my_vsnprintf ..................... Passed 0.04 sec
[30 Sep 2011 6:35]
Valeriy Kravchuk
With 5.5.17 (current mysql-5.5 tree) on Mac OS X I see no problems:
macbook-pro:mysql-5.5-work openxs$ make test
Running tests...
Test project /Users/openxs/bzr2/mysql-5.5-work
Start 1: pfs_instr_class
1/12 Test #1: pfs_instr_class .................. Passed 0.12 sec
Start 2: pfs_instr_class-oom
2/12 Test #2: pfs_instr_class-oom .............. Passed 0.09 sec
Start 3: pfs_instr
3/12 Test #3: pfs_instr ........................ Passed 0.03 sec
Start 4: pfs_instr-oom
4/12 Test #4: pfs_instr-oom .................... Passed 0.03 sec
Start 5: pfs
5/12 Test #5: pfs .............................. Passed 0.67 sec
Start 6: bitmap
6/12 Test #6: bitmap ........................... Passed 5.14 sec
Start 7: base64
7/12 Test #7: base64 ........................... Passed 0.10 sec
Start 8: my_vsnprintf
8/12 Test #8: my_vsnprintf ..................... Passed 0.07 sec
Start 9: my_atomic
9/12 Test #9: my_atomic ........................ Passed 2.14 sec
Start 10: my_rdtsc
10/12 Test #10: my_rdtsc ......................... Passed 0.16 sec
Start 11: lf
11/12 Test #11: lf ............................... Passed 2.29 sec
Start 12: my_malloc
12/12 Test #12: my_malloc ........................ Passed 0.05 sec
100% tests passed, 0 tests failed out of 12
Total Test time (real) = 11.08 sec
[30 Dec 2011 18:30]
Valeriy Kravchuk
Please, check if the problem is still repeatable in your environment with recent version, 5.5.19.
[4 Jan 2012 13:17]
Oleksii Moskvin
Checked. Works fine with gcc but still fails the same way using latest version of Intel compiler (icc).
[23 Jan 2012 19:07]
Sveta Smirnova
Thank you for the feedback. Verified as described.
[4 Feb 2013 6:11]
Sayantan Dutta
Not a bug. The code is supposed to work the way it is with gcc
[25 Feb 2013 13:13]
Sayantan Dutta
Bug Fixed.
[15 May 2013 15:11]
Paul DuBois
Test suite changes. No changelog entry needed.
[16 May 2013 13:46]
Laurent Blume
I am hitting that bug on Solaris 10, compiling MySQL 5.5.31 with Studio 12.3.
This is most definitely a bug: one of the tests is disabled in my_vsnprintf-t.c if the compiler is not GCC (#if defined (__GNUC__) clause), but the number of tests expected to pass is kept the same.
So it's impossible that those tests succeed if the compiler is not GCC, which makes me wonder how there is an official build for Solaris that doesn't even pass "make test".
I've made that change, it is enough for make test to succeed for me:
--- my_vsnprintf-t.c.original lun. mars 25 14:14:58 2013
+++ my_vsnprintf-t.c jeu. mai 16 15:35:30 2013
@@ -31,7 +31,11 @@
int main(void)
{
+#if defined (__GNUC__)
plan(58);
+#else
+ plan(57);
+#endif
test1("Constant string",
"Constant string");
[3 Jun 2013 14:09]
Paul DuBois
Changes for test suite. No changelog entry needed.
Description: When running "make test" in source folder after compilation tests always fail on my_vsnprintf checks [root@srv mysql-5.5.16]# make test Running tests... Start processing tests Test project /usr/local/src/mysql-5.5.16 1/ 12 Testing pfs_instr_class Passed 2/ 12 Testing pfs_instr_class-oom Passed 3/ 12 Testing pfs_instr Passed 4/ 12 Testing pfs_instr-oom Passed 5/ 12 Testing pfs Passed 6/ 12 Testing bitmap Passed 7/ 12 Testing base64 Passed 8/ 12 Testing my_vsnprintf *** Failed 9/ 12 Testing my_atomic Passed 10/ 12 Testing my_rdtsc Passed 11/ 12 Testing lf Passed 12/ 12 Testing my_malloc Passed How to repeat: cmake . make make test Suggested fix: in ./unittest/mysys/my_vsnprintf-t.c following function tests my_vsnprintf. 1-st argument is the expected result which my_vsnprintf should produce by processing 2..n args. 22 void test1(const char *res, const char *fmt, ...) 23 { 24 va_list args; 25 size_t len; 26 va_start(args,fmt); 27 len= my_vsnprintf(buf, sizeof(buf)-1, fmt, args); 28 va_end(args); 29 ok(strlen(res) == len && strcmp(buf, res) == 0, "\"%s\"", buf); 30 } and here we have statement which fails 157 test1("Hello string `I am a string`", 158 "Hello string %`s", "I am a string"); %`s seems to be a problem - backquote in invalid place and 2-nd backquote is missing. Suggested fix is 157 test1("Hello string `I am a string`", 158 "Hello string `%s`", "I am a string");