Changes for 2.4.1 for i686 compatibility #247
Conversation
On systems such as i686 with 32-bit long integers, expressions such as 60L * 60 * 1000 * 1000 (3600000000) overflow the maximum value of 2147483647, causing the test to fail. This can be fixed by using a "long long" (which is used for intmax_t) instead, e.g. 60LL * 60 * 1000 * 1000
src/lib-imap/test-imap-date.c
Outdated
| @@ -25,7 +25,7 @@ static void test_imap_date(void) | |||
| { "31-Dec-9999", 253402214400LL }, | |||
| #endif | |||
| /* conversions to maximum values */ | |||
| #if TIME_T_MAX_BITS == 31 | |||
| #if TIME_T_MAX_BITS == 31 || (TIME_T_MAX_BITS == 32 && defined(TIME_T_SIGNED)) | |||
There was a problem hiding this comment.
Hmm. I think these checks are assuming TIME_T_MAX_BITS=31 with signed 32bit time_t. Also m4/gmtime_max.m4 has:
printf "check failed, assuming "
i_cv_gmtime_max_time_t=31
So I'm thinking it might be better to fix the m4 test to output TIME_T_MAX_BITS=31 instead with 32bit signed time_t. Then these other checks wouldn't need to change (except tm_is_too_large() probably).
There was a problem hiding this comment.
Although I guess that would break utc_mktime(), and there are other checks assuming TIME_T_MAX_BITS=32. Maybe just i_cv_gmtime_max_time_t=31 should be changed to 32.
There was a problem hiding this comment.
Also in this commented case TIME_T_MAX_BITS == 31 would never happen, so I guess that should be simply #if #if TIME_T_MAX_BITS == 31 || (TIME_T_MAX_BITS == 32 && defined(TIME_T_SIGNED)
Or no, maybe it's nice to have the 31 check after all, even if it won't practically happen..
The code and tests have some hard-coded values based on the size of time_t but they don't take into account whether or not time_t is a signed value. On systems such as i686 with time_t as a signed 32-bit integer, the maximum value is the same as an unsigned 31-bit integer. Tweak the configure test to treat 32-bit signed as 31 bits.
|
Changed the m4 test to make it return 31 bits instead of 32 if time_t was signed, and that made most of the rest of the changes redundant. |
This PR changes a few files (mainly test files) to get the test suite to pass with linux on i686. This reflects some filter values not fitting in an i686 long integer (32 bits), and fixes for a signed 32-bit time_t. The configure script already does all of the necessary checks; it's just a case of using the results properly.