The AIX libc glob is a bit anemic feature wise (AFAIK missing a few flags), and I'm suspecting it may be buggy. I've had a few complaints about it.

I was thinking of vendoring say, a BSD's implementation in our application (and am tempted to do just that), but I bet it'd probably be better if it were provided in here, since other software may be using it.

I noticed that we have backtrace_symbols I think we should also have backtrace_symbols_fd.

This will be handy when building open source projects.

For example when porting over mariadb I ran into an into a linker error because bactrace_symbols_fd is not implemented.

For compatibility, we need to map EACCES to EWOULDBLOCK in libutil_flock when LOCK_NB is set.

https://www.man7.org/linux/man-pages/man3/err.3.html

Would be useful to have a password change utility, like the standard Unix passwd.

The easiest way to implement such a tool would be to use QSYCHGPW using _PGMCALL.

Helpfully, the API takes in CCSID arguments for both old and new passwords, so we can pass in 1208 (or the output of Qp2paseCCSID).

When run as root, passwd doesn't prompt for the old password. The API does support this by passing *NOPWD, for the old password, but the user would need *SECADM special authority. We could check for this with #11 then not prompt.

dladdr is a GNU extension to the common dl family of functions: dlopen, dlsym, etc.

It has been implemented on many other operating systems:

• Solaris
• FreeBSD
• macOS
• HP-UX

It seems like the FreeBSD and macOS implementations are ports of or just compatible with the Solaris version.

One thing we need to think about is what to return for the dli_fname? Do we include the member name (if any) or just the file path. A given path that includes the member in the /path/to/lib(member.so) syntax is a valid file path, so that could be confusing, however it would be unique to AIX/PASE. The HP-UX version at least has defined system-specific fields, so I think we can follow them.

no check for the failed malloc in line 83. when buffer is dereferenced, will cauce null deference

asprintf.h was consolidated to stdio.h but is still referenced in the makefile

I think it should work something like this:

// sort of modeled after getpwnam/getpwuid
int get_special_authorities_uid(uid_t uid);
int get_special_authorities_nam(const char* name);

int auth = get_special_authorities_uid(0); // QSECOFR
assert(auth & AUTH_ALLOBJ);

int auth = get_special_authorities("*CURRENT");
if (auth & AUTH_JOBCTL) printf("Current user has *JOBCTL\n");

The API would be based on QSYUSRI.

@ThePrez had a previous attempt when it was on BitBucket: https://bitbucket.org/litmis/portlibfori/pull-requests/1/checkspcaut/diff That went a different way, using QSYCUSRS instead. The semantics are slightly different between these two approaches.

QSYUSRI tells which authorities are given to a user profile object outright, but an MI program can adopt or drop authorites so the authorities checked by QSYCUSRS could be different for the same user when called by different programs. PASE does not support adopted authority, though, so it's largely irrelevant unless you're calling in to PASE from ILE, which isn't very common. For any program started from PASE, it will only have the authorities specified by the user profile that it was started under.

This is needed by util-linux

https://man7.org/linux/man-pages/man3/asprintf.3.html
https://www.freebsd.org/cgi/man.cgi?query=asprintf&apropos=0&sektion=3&manpath=FreeBSD+6.2-RELEASE&format=html

It can be created by wrapping vsnprintf, eg something like this but with error checking:

int size = vsnprintf(NULL, 0, fmt, args);
char* str = malloc(size+1);
vsnprintf(str, size+1, fmt, args);

While running mysql_install_db I ran into the following error:

/QOpenSys/pkgs/bin/mysql_install_db[2]: 8628721 Memory fault(coredump)

Examining the core file I found:

Program terminated with signal SIGSEGV, Segmentation fault.
#0  0x0900000000985dc0 in backtrace () from /QOpenSys/pkgs/lib/libutil.so.2(shr_64.o)
(gdb) bt
#0  0x0900000000985dc0 in backtrace () from /QOpenSys/pkgs/lib/libutil.so.2(shr_64.o)
#1  0x0000000100031c48 in sf_malloc (size=1480, my_flags=16)
    at /home/AMUSSE/rpmbuild/BUILD/mariadb-10.3.14/mysys/safemalloc.c:141
#2  0x000000010003074c in my_malloc (size=1480, my_flags=16)
    at /home/AMUSSE/rpmbuild/BUILD/mariadb-10.3.14/mysys/my_malloc.c:101
#3  0x0000000100f042ec in init_queue (queue=0x180cee7e8 <_thralarm.bss_3_+312>, max_elements=184, 
    offset_to_key=0, max_at_top=0 '�', compare=@0x1803cc4d8: 0x10129fe08 <compare_ulong>, first_cmp_arg=0x0, 
    offset_to_queue_pos=25, auto_extent=10) at /home/AMUSSE/rpmbuild/BUILD/mariadb-10.3.14/mysys/queues.c:79
#4  0x000000010129ff00 in init_thr_alarm (max_alarms=183)
    at /home/AMUSSE/rpmbuild/BUILD/mariadb-10.3.14/mysys/thr_alarm.c:92
#5  0x0000000100009dec in signal_hand (arg=0x0) at /home/AMUSSE/rpmbuild/BUILD/mariadb-10.3.14/sql/mysqld.cc:3512
#6  0x0900000000583da4 in _pthread_body () from /QOpenSys/usr/lib/libpthreads.a(shr_xpg5_64.o)
#7  0x0000000000000000 in ?? ()

https://linux.die.net/man/3/warnx

The err() and warn() family of functions display a formatted error message on the standard error output. ... The output is terminated by a newline character.

https://github.com/IBM/portlibfori/blob/master/util/getopt_long.c#L157

This warnx macro implementation does not produce a newline.

Functions like openat are broken under PASE (is this documented anywhere?), where they only work on AT_FDCWD instead of arbitrary FDs, returning EINVAL. This has made a lot of people very angry and been widely regarded as a bad move, because most at function users are using non-cwd paths.

It is possible to write a wrapper. It would get the current working directory as an FD, switch CWD to it so that the FD-relative path would work, then switch back once the non-at version of the function has been called. Note that open is annoying because of it using variadic arguments, but many other functions outside of say, linkat could perhaps be generated from a template. Also note there's no mutex, so you will have a very bad time in a threaded application.

int openat_shim(int fd, const char *path, int flags, ...)
{
        // For the shim, get current cwd, change current cwd to passed fd, open file, then change back.
        char current_cwd[MAXPATHLEN];
        getcwd(current_cwd, MAXPATHLEN);
        int current_cwd_fd = open(current_cwd, O_RDONLY);
        if (current_cwd_fd == -1) {
                return -1;
        }
        if (fchdir(fd) == -1) {
                goto fin;
        }
        // Massage the va_list if required
        int retval = -1;
        if (flags & O_CREAT) {
                mode_t mode;
                va_list ap;
                va_start(ap, flags);
                mode = va_arg(ap, mode_t);
                retval = open(path, flags, mode);
                va_end(ap);
        } else {
                retval = open(path, flags);
        }
fin:
        fchdir(current_cwd_fd);
        close(current_cwd_fd);
        return retval;
}

Ideally, it'd be nice if PASE had this fixed properly, but 🤷

This is due to a problem when the wrapper header finds the wrapper header as the next in the include_next chain.

gcc -c -D__PASE__ -fPIC -Iutil -isystem /QOpenSys/usr/include -o util/getopt_long.o util/getopt_long.c
util/getopt_long.c: In function 'pase_warn':
util/getopt_long.c:172:10: warning: implicit declaration of function 'vfprintf' [-Wimplicit-function-declaration]
     rc = vfprintf(stderr, fmt, args);
          ^~~~~~~~

Here util/getopt_long.c has #include <stdio.h>. Because of -Iutil`, this gets loaded from util/stdio.h which has

#ifndef LIBUTIL_SYS_FILE_H
#define LIBUTIL_SYS_FILE_H

#include_next <stdio.h>
...

Because of our GCC in PASE injects -I/QOpenSys/pkgs/include, if this wrapper header is installed, it will find this one next. Now when it gets loaded, because the LIBUTIL_SYS_FILE_H header guard is already defined the #ifndef block will be skipped and the real system stdio.h is never included.