Piero Bosio Social Web Site Personale

@nabijaczleweli Solaris itself actually only ships gcc & clang because the Studio compilers are unbundled extra cost addons that are in Extended Support now (until at least next year) - it is true the kernel and most of the core OS are built with them still though.

@nabijaczleweli our C library & headers are up to date for C11 - work on C23 is on my long-term todo list, but I've gotten to it yet.

@nabijaczleweli Solaris 11.4.90 includes packages for gcc 15 and clang 21, and many of our FOSS packages are built with those (mostly gcc).

My other contributions to this SRU included removing gdk-pixbuf-xlib, gtk-vnc, and libneon; converting more of our X11 packages to build with gcc; updating FreeType to 2.14.1; marking the Imake package as obsolete and planned to be removed in a future SRU; and fixing bugs in the Xserver and gdm.

Diffs of the core OS man pages between #OracleSolaris 11.4.87 and 11.4.90 are at https://gist.github.com/alanc/cfbaa19eb1ee04afe996da97f121fe7e so you can see more about what changed in the software, as well as some of the documentation cleanups we did.

My big man page project for the quarter was a cleanup pass across the section 4m (kernel module) man pages.

Work continued on trying to prevent bugs that result in killing all processes on the system by mistake, in response to customers hitting https://github.com/sudo-project/sudo/issues/458 last year, and bugs in other programs using -1 as a pid value in kill() calls by mistake.

Previously, in #OracleSolaris 11.4.83, killpg() was made to reject the always invalid process group id of 1, instead of turning it into the "kill all processes" argument of -1 to kill().

Now in 11.4.90, the kill system call was modified to not allow processes to use a pid of -1 unless they'd specifically set a process flag that they intend to kill all processes first, to help with programs that didn't check for errors when finding the process id for the singular process they wanted to kill.

New utime32_t and utimeval32 types have been defined in the system headers in #OracleSolaris 11.4.90 for places where binary compatibility forces a 32-bit time value, but unsigned values can be used to represent times from 1970-2106 instead of the traditional signed 32-bit range of 1901-2038.

The bufmod(4m) interface was modified to use these, allowing snoop(8) to become #Y2038 safe (though still destined to fail in 2106). snoop also became a 64-bit program in this release.

#beadm gained the ability to set a "preserve" property in #OracleSolaris 11.4.90 to mark a boot environment as not to be destroyed by either automation or manual "beadm destroy" commands.

#ZFS changes in #OracleSolaris 11.4.90 include more flexibility in setting retention properties when receiving a new file system, and adding the ability for zfs scrub and resilver to run before all the blocks have been freed from previous zfs destroy operations. (This requires upgrading pools to the new zpool version 54.)

Today we released #OracleSolaris 11.4.90, our quarterly feature update. The announcement is posted to the blog at https://blogs.oracle.com/solaris/announcing-oracle-solaris-11-4-sru90
and a more detailed What's New blog post is in the pipeline to be published soon.