Big Surprise: Future 2.5 kernel releases will use a forward port of 2.4's IDE core (current BK kernels already do); and Marcin Dalecki will no longer be IDE maintainer for 2.5.

It is unclear who will take his place, but Linus writes: "Alan seems to be not completely unwilling to work on it, and unlike me he _can_ interact with Andre most of the time. Possibly Jens will do the 2.5.x side, of it (with Alan working on 2.4), but we've not talked it through".

GCC 3.2 is out. The focus for this release is the C++ ABI and some small bugfixes for the C++ standard library. No significant changes have been made to the other compilers (C, Objective-C, Java, Ada and Fortran) since the 3.1.1 release a few weeks ago. You might want to read this warning about C++ ABI compatability, if you plan on installing GCC 3.2 (mirrors).

FreeBSD 4.6.2 has been released. This is a fixpoint release that resolves several ATA-issues and addresses security problems such as the recent sshd, openssl and zlib vulnerabilities, aswell as some bugs in the syncookie and resolver code. A more complete list of changes can be found in the release notes (i386, alpha) on the FreeBSD website.

You can download it from the official FreeBSD ftp server, or even better (and faster) from one of the mirrors.

For those that are missing pre-patches: Jeff Garzik announced he had "created an automated BK snapshot process to make BK changes accessible between kernel releases"; since Linus is not doing pre-patches anymore.

They are available at:
ftp://ftp.kernel.org/pub/linux/kernel/people/jgarzik/snap/2.5/

All BK changes not in the current 2.5 kernel release will be posted at this URL, in GNU patch form, on a nightly basis.

Andreas Gruenbacher posted a patch to lkml that would allow for "dynamically sized caches that shrink under pressure" (caches that shrink when available physical memory is low). This feature seems a bit controversial, since the traditional approach has always been to cache as much as possible, and to swapout under pressure. The rationale is that you want to have as much cache in RAM, to reduce disk access.

Members of the other party, however, argue that having too much cache RAM is counterproductive; and that it is better to free unused cache under pressure.

It looks like this patch will make it in, though. Quoth Linus: "Yah, it's a guess. We can test it."

Update (2002-08-13): Corrected the "too much cache on disk" sentence. Thanks to Hans Reiser for spotting it :)

The GNU/Hurd J1 images have been built and are available online at ftp://ftp.gnu.org/iso/hurd-J1.

The full announcement follows.

An Anonymous source wrote: Linux 2.4.19-rc5 was released as 2.4.19 in a new (to LK development) era in which the last release candidate is relabeled the official release. This process provides a much higher degree of security and peace-of-mind in the release kernels, which have been known to have with typos and compile failures and even data corruption bugs.

While I agree that the new release method is better, I am curious how many users use a newly released stock kernel. Most of the users I know use heavily patched vendor kernel. So does this change really matter?

In this article at Linux Journal, Julian Rose explains dsp_K, an "open-source, real-time kernel and library for digital signal processors (DSPs)". It covers both technical and historical ground, paying particular attention to the board support package (BSP) layer that is at the kernel's heart; and explains how dsp_K handles funcionality that is essential in any kernel, like context switching, and interrupt handling.

The BSP comprises the minimal services required to provide a functional software platform for an application to run, along with a minimal kernel personality, interrupts, context switching, scheduler sanity checker and task functions

The Hurd currently runs only on the Mach microkernel, but there is work to port it to the newer L4 microkernel. This article has the details about the current status of the Hurd/L4 port. It is primarily written for people familiar with Hurd/Mach, but who do not know L4.

Because of design differences between L4 and Mach (as is to be expected from 2 different kernels), a lot of infrastructural work still needs to be done; including a Virtual Memory Manager (VMM), getting glibc ported, and writing device drivers.

Everyone's favorite Compiler Collection, GCC, has been upgraded to version 3.1.1. GCC 3.1.1 is a bugfix release, no new features have been introduced.

According to Mark Mitchell (the Release Manager), GCC 3.2 "will be available very soon (within days or a week). The only changes in GCC 3.2 relative to GCC 3.1.1 will be changes to the C++ ABI."

Major fixes include (taken from the Changes page):

• A bug related to how structures and unions are returned has been fixed for powerpc-*-netbsd*.
• An important bug in the implementation of -fprefetch-loop-arrays has been fixed.
  Previously the optimization prefetched random blocks of memory for most targets except for i386.
• The Java compiler now compiles Java programs much faster and also works with parallel make.
  
• Nested functions have been fixed for mips*-*-netbsd*.
  
• Some missing floating point support routines have beed added for mips*-*-netbsd*.
  
• This message gives additional information about the bugs fixed in this release.

Sometimes kernel functionality must be accessed by userspace applications. This causes problems, because some aspects of the kernel ABI are only available through the include files. The problem is that the kernel headers include a lot of other kernel headers, which contain stuff not suitable for userspace. (This is not about system calls, but other aspects of the kernel).

So, H. Peter Anvin proposed having a set of "common ABI description files" that would be shared by userspace and the kernel; since he would be working on a "very tiny libc for initramfs, and will likely have to deal withhow to use the kernel ABI as well".

Andreas Schuldei asked in lkml if there was an easy a way for BitKeeper to backport stuff from the 2.5 to the 2.4 repositories. Like so many discussions, this rapidly veered off; from keeping separate BK repositories in sync, to the limitations of BitKeeper.

Much of the resulting discussion follows.

NetBSD (the most widely-ported OS in the world) 1.5.3 is finally out! From the Recent Changes page:

NetBSD 1.5.3, a patch release improving stability, fixing bugs in, addressing security issues in, and adding some features to NetBSD 1.5.2, has been released with support for 20 architectures. More information is available in the 1.5.3 release announcement.

Linus has integrated the reverse page mapping infrastructure of Rik van Riel's virtual memory system, using Andrew Morton's [interview] minimal-rmap patch [story].

The patch implements the reverse mapping system on top of Andrea's VM system, but should allow enhancements to the virtual memory system especially with respect to page replacement (swapping) such as better performance, and predictability under load as well as more accurate accounting.