Not entirely portable use of usleep() in tests

[rsbeckerca]

I did this

Compiled curl with OpenSSL and c99 on NonStop x86, and got a compile error on usleep(100*1000). This function is only available in threaded code on NonStop. I need to use sleep(100) instead if not threaded.

Any objection if I put a pull request together for this?

I expected the following

Clean compile of the test suite.

curl/libcurl version

curl 8.11.0

operating system

NonStop L24.08

[vszakats]

Any reason not to use sleep() on all platforms? (Except on Windows, which needs it remapped to Sleep())

[rsbeckerca]

Any reason not to use sleep() on all platforms? (Except on Windows, which needs it remapped to Sleep())

I would have though that the technique used for Sleep() throughout the code would be better, but I'm somewhat confused as to how that even works. However, sleep() on NonStop does not appear to be thread blocking in all threading models, while usleep() has a safe wrapper. This is only in ws-data.c and ws-pingpong.c.

[rsbeckerca]

My proposed PR is now up.

[bagder]

Any reason not to use sleep() on all platforms? (Except on Windows, which needs it remapped to Sleep())

Because POSIX sleep() is in number of seconds, usleep is microseconds. Apparently NonStop has a special sleep().

[vszakats]

Any reason not to use sleep() on all platforms? (Except on Windows, which needs it remapped to Sleep())

Because POSIX sleep() is in number of seconds, usleep is microseconds. Apparently NonStop has a special sleep().

Ah, micro vs. milli got me!

[rsbeckerca]

Any reason not to use sleep() on all platforms? (Except on Windows, which needs it remapped to Sleep())

Because POSIX sleep() is in number of seconds, usleep is microseconds. Apparently NonStop has a special sleep().

Ah, micro vs. milli got me!

Actually I meant milli, but sleep() is seconds.