Don't include sys/select.h on HP-UX as it doesn't exist

[tomhughes]

There is no sys/select.h on HP-UX and the man page lists sys/time.h as the correct header for select.

[mattcaswell]

I'm wondering why we haven't hit this before on this platform?

[rsbeckerca]

I'm wondering why we haven't hit this before on this platform?

We have the same situation on HPE NonStop. sys/time.h is where select is declared also. I recall encountering it during our report.

[tomhughes]

That code is new in 3.x I think - the only place in 1.1 that includes it is include/internal/sockets.h which does this:

#  ifdef OPENSSL_SYS_AIX
#   include <sys/select.h>
#  endif

which is only used on AIX systems.

[rsbeckerca]

Is this on master or openssl-3.0? We just built 3.0 last night without errors.

[tomhughes]

This came from me building the 3.0.3 release but looking at it I can see our Itanium machine (running B.11.23) does have it - it's only the PA-RISC machine (running B.11.11) which doesn't.

Both build without including it though.

[rsbeckerca]

This came from me building the 3.0.3 release but looking at it I can see our Itanium machine (running B.11.23) does have it - it's only the PA-RISC machine (running B.11.11) which doesn't.

Both build without including it though.

I'm rerunning the builds on our ia64 also. The x86 build 3.0 was fine last night.

[mattcaswell]

Is this on master or openssl-3.0? We just built 3.0 last night without errors.

There is specific code for NonStop to avoid this:

# if defined __TANDEM
#  include <unistd.h>
#  include <sys/time.h> /* select */
#  if defined(OPENSSL_TANDEM_FLOSS)
#   include <floss.h(floss_select)>
#  endif
# elif defined _WIN32
#  include <winsock.h> /* for type fd_set */
# else
#  include <unistd.h>
#  if defined __VMS
#   include <sys/socket.h>
#  else
#   include <sys/select.h>
#  endif
# endif

[rsbeckerca]

Is this on master or openssl-3.0? We just built 3.0 last night without errors.

There is specific code for NonStop to avoid this:

# if defined __TANDEM
#  include <unistd.h>
#  include <sys/time.h> /* select */
#  if defined(OPENSSL_TANDEM_FLOSS)
#   include <floss.h(floss_select)>
#  endif
# elif defined _WIN32
#  include <winsock.h> /* for type fd_set */
# else
#  include <unistd.h>
#  if defined __VMS
#   include <sys/socket.h>
#  else
#   include <sys/select.h>
#  endif
# endif

Yup. That looks like my handiwork ;)

[rsbeckerca]

On the other hand, @mattcaswell, we are getting

make: *** No rule to make target 'include/openssl/configuration.h', needed by 'build_generated'.  Stop.

as of this morning.

[rsbeckerca]

On the other hand, @mattcaswell, we are getting

make: *** No rule to make target 'include/openssl/configuration.h', needed by 'build_generated'.  Stop.

as of this morning.

That one results from make clean removing configuration.h

[mattcaswell]

That one results from make clean removing configuration.h

Ah. Can you raise a separate issue for that please?

[rsbeckerca]

That one results from make clean removing configuration.h

Ah. Can you raise a separate issue for that please?

Will do.

[rsbeckerca]

Opened as #18396

[openssl-machine]

24 hours has passed since 'approval: done' was set, but as this PR has been updated in that time the label 'approval: ready to merge' is not being automatically set. Please review the updates and set the label manually.

[t8m]

I've merged this to master and 3.0 branches. I did not notice this has CLA:trivial before merging this but I am OK with it. @mattcaswell Can you please confirm, you're OK with CLA: trivial for this?

[rsbeckerca]

This builds correctly on both master and openssl-3.0 on NonStop x86 and ia64