Refactor use of size_t #602
Conversation
Use uint64_t in preference to size_t to avoid size mismatches on architectures that are not 64bit. This generates compilation errors on most modern gcc compiler version. Signed-off-by: James Page <james.page@ubuntu.com>
| @@ -86,7 +86,7 @@ void tcmur_cmd_complete(struct tcmu_device *dev, void *data, int rc) | |||
| } | |||
|
|
|||
| static void tcmur_cmd_iovec_reset(struct tcmur_cmd *tcmur_cmd, | |||
| size_t data_length) | |||
| uint64_t data_length) | |||
There was a problem hiding this comment.
size_t is 32 bits on 32bit archs, right? If so, above we will use a uint64_t, but the iov_len field on a struct iovec is size_t, so will that cause a compilation error?
There was a problem hiding this comment.
Hmm - it does not create a compile error but I think that the code needs alignment around this fact - i.e size_t should be the basic assumption rather than uint64_t
|
Hey, thanks for the patch. One other question I had is are using tcmu-runner on a 32bit arch or just handling build issues? We had another 32 bit issue here: so I am just wondering if you also hit that. |
|
I'm mainly just dealing with build issues in Ubuntu; the packaging had disabled errors relating to type mismatches due to size_t != uint64_t on 32 bit architectures. I had not hit the issue in #360. |
Use uint64_t in preference to size_t to avoid size mismatches
on architectures that are not 64bit. This generates compilation
errors on most modern gcc compiler version.
Signed-off-by: James Page james.page@ubuntu.com