USE_SOLARIS for CMake conversion #74

Basic information

TDE version: N/A
Distribution: Solaris
Hardware: N/A

Description

With autoconf/automake, USE_SOLARIS is defined as CPPFLAGS for Solaris:
https://mirror.git.trinitydesktop.org/gitea/TDE/tde-common-admin/src/branch/master/acinclude.m4.in#L4341
But with CMake, it is not defined.

In CMake conversioned TDE modules, USE_SOLARIS is still used for Solaris support codes.

Then, Solaris support codes will not be used on Solaris.

Steps to reproduce

build TDE modules with CMake on Solaris
OS depending features will not work well

Screenshots

#ifdef USE_SOLARIS can be safely replaced with
#if (defined(sun) || defined(__sun)) && (defined(__SVR4) || defined(__svr4__))

But #ifdef __sun should be good enough.

Hi Obata-san,
could you please prepare a PR for this fix? Thanks a lot 😄

Can someone create the list affected modules?

I understand correctly that instead of adding the definition of USE_SOLARIS at the global level of TDE CMake macros, we will solve the use of #ifdef __sun in modules where it is needed?

Hi Obata-san,
maybe I misunderstood what you wanted to do.
I see USE_SOLARIS being used in admin module, when we build packages with autotools/automake. I don't see USE_SOLARIS in cmake-trinity, so I thought you intended to add support for it in USE_SOLARIS.
Re-reading your comment again, I now understand you are suggesting to replace USE_SOLARIS everywhere in the code with #if (defined(sun) || defined(__sun)) && (defined(__SVR4) || defined(__svr4__)).
Is that correct? If so, I can work on it in coming days but I won't be able to test. Would you like me to prepare PR for you to test each individual module or are you happy enough if I push to master/R14.0.x and you just rebuild?

I just found the issue, but not ready to test it, so created this ticket for the reminder (intend to be meta ticket).
So I'm happy enough with just changed codes.

defined(sun) is for ancient compiler(?), defined(__SVR4) || defined(__svr4__) is "is SVR4" i.e. "not for SunOS 4", then defined(__sun) should be used here.

Ok, thanks @obache. I will work on replacing #ifdef USE_SOLARIS with #ifdef __sun in coming days/weeks.

@obache could you please check TDE/smb4k#4 to make sure what I didn is ok for you?

In the code there is also _OS_SOLARIS_ and Q_OS_SOLARIS.
The first one is not defined anywhere, so I guess we can replace it with __sun as well. The second one is defined in TQt3 if __sun is defined. So the question is whether to leave Q_OS_SOLARIS as is or simply replace it with __sun.
What do you think? (@SlavekB @obache)

As far as I know, TQt3 defines Q_OS... to make it certainty that it is uniform and unambiguous. Therefore, a good idea seems to leave these definitions in TQt3. However, in other code it seems that these Q_<something> constants are used rarely. It is hard to decide whether to prioritize these constants specific to TQt or whether to stick to standard constants, although quite inconsistent.

Thanks @SlavekB. Over the night I thought it may be good to use Q_OS_* across the code. So leave the Q_OS_SOLARIS define in TQt3 and replace __sun/__solaris__ with Q_OS_SOLARIS in the other modules.
The reason for this is to have a unique central place where we detect the OS and use it uniformly across the rest of the code base.
What do you think?

Yes, the use of constants Q_OS_<something> seems like a good idea because the names of these constants are well consistent.