diff options
author | Albert Cheng <acheng@hdfgroup.org> | 1998-11-16 19:08:18 (GMT) |
---|---|---|
committer | Albert Cheng <acheng@hdfgroup.org> | 1998-11-16 19:08:18 (GMT) |
commit | e7643394b0a3d761b438ff22a7b6372d1182dd15 (patch) | |
tree | 3ac321e6731ba4150502cc0248bbcc871b7467fd /test/big.c | |
parent | df77c715ccec86c58b648ad780dc84f633a929b8 (diff) | |
download | hdf5-e7643394b0a3d761b438ff22a7b6372d1182dd15.zip hdf5-e7643394b0a3d761b438ff22a7b6372d1182dd15.tar.gz hdf5-e7643394b0a3d761b438ff22a7b6372d1182dd15.tar.bz2 |
[svn-r916] Problem:
In machine (e.g. O2K) where long long and long are the same sizes, it
ended up defining GB8LL as 1 which is way too small for the dataset size.
Solution:
Add an entry to define GB8LL if long can hold it. When all fails,
define GB8LL as 0. During runtime, enough_room() will detect this
special value and skips the test.
Platform tested:
O2K IRIX64.
Diffstat (limited to 'test/big.c')
-rw-r--r-- | test/big.c | 8 |
1 files changed, 7 insertions, 1 deletions
@@ -33,8 +33,10 @@ #define GB8LL ((unsigned long long)8*1024*1024*1024) #elif SIZEOF___INT64 > SIZEOF_LONG #define GB8LL ((unsigned __int64)8*1024*1024*1024) +#elif SIZEOF_LONG >= 8 +#define GB8LL ((unsigned long)8*1024*1024*1024) #else -#define GB8LL ((long)1) +#define GB8LL ((unsigned long)0) /* can not do the test */ #endif static hsize_t @@ -135,6 +137,10 @@ enough_room(void) size_t i, size = (size_t)1 << 30; char name[32]; + /* verify if this machine supports 8GB sizes */ + if (GB8LL==0) + goto done; + /* Initialize file descriptors */ for (i=0; i<NELMTS(fd); i++) fd[i] = -1; |