Add Solaris build

[psumbera]

No description provided.

[0-wiz-0]

Thank you for the workflow!
As for the CMake part: in the Solaris docs at https://docs.oracle.com/cd/E88353_01/html/E37843/memset-s-3c.html it is claimed that #define __STDC_WANT_LIB_EXT1__ 1 is needed, and we already do that.
What Solaris version are you using exactly?

[psumbera]

Thank you for the workflow! As for the CMake part: in the Solaris docs at https://docs.oracle.com/cd/E88353_01/html/E37843/memset-s-3c.html it is claimed that #define __STDC_WANT_LIB_EXT1__ 1 is needed, and we already do that. What Solaris version are you using exactly?

You are right. But in case of lib/zip_buffer.c it's defined after inclusion of <string.h>.

Not sure what is proper solution.

[0-wiz-0]

I thought I had fixed that in e16526d - which version are you testing against?

[psumbera]

Ok, I saw the problem in 1.11.4. I didn't verify latest commits. So hopefully it does work now.

[0-wiz-0]

Thank you, merged!

[0-wiz-0]

Hm, ziptool_regress fails to build with

[ 91%] Building C object regress/CMakeFiles/ziptool_regress.dir/ziptool_regress.c.o
  In file included from /home/runner/work/libzip/libzip/src/ziptool.c:34,
                   from /home/runner/work/libzip/libzip/regress/ziptool_regress.c:94:
  /home/runner/work/libzip/libzip/src/ziptool.c: In function ‘zstat’:
  /home/runner/work/libzip/libzip/lib/compat.h:189:23: error: implicit declaration of function ‘localtime_s’; did you mean ‘localtime_r’? [-Wimplicit-function-declaration]
    189 | #define zip_localtime localtime_s
        |                       ^~~~~~~~~~~
  /home/runner/work/libzip/libzip/src/ziptool.c:675:15: note: in expansion of macro ‘zip_localtime’
    675 |         tpm = zip_localtime(&sb.mtime, &tm);
        |               ^~~~~~~~~~~~~
  /home/runner/work/libzip/libzip/src/ziptool.c:675:13: error: assignment to ‘struct tm *’ from ‘int’ makes pointer from integer without a cast [-Wint-conversion]
    675 |         tpm = zip_localtime(&sb.mtime, &tm);
        |             ^
  gmake[2]: *** [regress/CMakeFiles/ziptool_regress.dir/build.make:76: regress/CMakeFiles/ziptool_regress.dir/ziptool_regress.c.o] Error 1
  gmake[1]: *** [CMakeFiles/Makefile2:1484: regress/CMakeFiles/ziptool_regress.dir/all] Error 2
  gmake: *** [Makefile:146: all] Error 2
[185](https://github.com/nih-at/libzip/actions/runs/21400471581/job/61609832772#step:3:1861)

[0-wiz-0]

https://docs.oracle.com/cd/E88353_01/html/E37843/localtime-s-3c.html again says that #define __STDC_WANT_LIB_EXT1__ 1 should fix this, and src/ziptool.c includes compat.h, which starts with

/* to have *_MAX definitions for all types when compiling with g++ */
 #define __STDC_LIMIT_MACROS

 /* to have ISO C secure library functions */
 #define __STDC_WANT_LIB_EXT1__ 1

Not sure yet what the problem is.

[psumbera]

Sorry, for the problems! Originally it was passing. I think it would be better to allow CI builds for all pull requests.

[psumbera]

I think that this resolves the issue:

diff --git a/regress/ziptool_regress.c b/regress/ziptool_regress.c
index 1cfd2921..00f3765a 100644
--- a/regress/ziptool_regress.c
+++ b/regress/ziptool_regress.c
@@ -1,3 +1,5 @@
+#include "compat.h"
+
 #include "zip.h"
 
 #include <sys/stat.h>

But overall I think that it might be better to define -D__STDC_WANT_LIB_EXT1__=1 for build (set(CMAKE_REQUIRED_DEFINITIONS "-D__STDC_WANT_LIB_EXT1__=1")) than having it in lib/compat.h.

[0-wiz-0]

I've fixed the Solaris build, thanks for the pointer.

As for your other suggestion - this is already set in the CMake file

./CMakeLists.txt:list(APPEND CMAKE_REQUIRED_DEFINITIONS -D__STDC_WANT_LIB_EXT1__=1)

so it should affect the whole build, and I'm not sure why it didn't affect ziptool_regress