Run without itimer support in alpine

[salsal97]

Added changes to -

1. Build on alpine
2. Run without timer support in the kernel using sleep in a separate thread and a signal to callback.

[swx-jenkins2]

Can one of the admins verify this patch?

[ChrisCoe]

Why is stdio.h being included to vma-redirect.h when vma-redirect.cpp has not been touched? Not clear to me.

[salsal97]

Same here, this is another change required to build sockperf in alpine OS

[ChrisCoe]

Same here, I see that you use os_thread_t in file client.cpp. Why not include header os_abstract.h in client.h?

[salsal97]

This is one of the changes required to properly build in alpine OS, the program was not able to find this file

[igor-ivanov]

Hi @salsal97, could you provide a motivation not to use setitimer

[salsal97]

@igor-ivanov , I am running this code inside an Intel SGX secure enclave that does not have support for the itimers syscall yet, hence these modifications in the upstream help us

[igor-ivanov]

Windows system does not have itimerval too.
Please consider changes in os_abstraction.h|cpp to support Alpine as it is done for WIN.
Probably it is possible to use something as

#ifdef ALPINE 

[salsal97]

@igor-ivanov, the changes made to client.cpp support the use of intel sgx enclaves, which might have itimer support as well in the future
the additions in terms of headers to the other two files were what were needed to build the project in alpine. How would one source the header file additions from os_abstract.cpp?

[igor-ivanov]

bot:retest