WizenedEE:
Why not just use a newer version of the toolchain?
There are at least 2 very good reasons:
1: Many incompatible changes have been added in newer versions. Several important details, especially regarding PROGMEM, have changed. Programs written with the old compiler often fail to compile, because toolchain changes were made without backwards compatibility. A toolchain upgrade will cause a lot of pain, by breaking many libraries and sketches.
2: Somebody would need to go through the work to produce 4 binary packages. Before you point to availability of the newer toolchain from various sources, consider the situation with the old toolchain on Linux. For many months, perhaps even a year or more, the Arduino developers were looking for someone to merely package the old toolchain at a .tgz file. I finally did it (with some certain libraries added and scripting tricks used so it's a self-contained package that runs on all Linux distros). Nobody else even offered. It's not extremely difficult, but someone needs to do the work, yet every time this subject comes up, people who argue for a newer toolchain point to various distros and binary packages that aren't quite the same version (rather than doing the harder job of actually creating 4 identical packages that are known to be the same). It's easy to say "use a newer toolchain", but creating the 4 actual .zip/.tgz files to be used on all Arduino IDEs is actual work.
WizenedEE:
The reason I've heard against updating is that this one works fine... but obviously it doesn't.
One small bug suddenly becomes the toolchain obviously doesn't work!
Until now, or more specifically a few days ago, there were no known problems with the malloc() version I've been using with Teensy. So far, this is the first time anybody has missed the __malloc_heap_end feature (and possibly other stuff... still waiting for Bill to respond specifically on the RTOS needs before I work on the code).
This is really a very minor issue. It'll probably take less time to fix that I've spent just writing these messages here and on the PJRC forum.
A toolchain upgrade is a lot of work. Anyone who really wants a toolchain upgrade can approach the Arduino developers. Maybe they'll go for it? One way that will NOT work is just suggesting they do all the work. Actually creating the 4 packages, testing them carefully across all many version of all 4 platforms, and against every bundled library and perhaps a couple dozen of the widely used 3rd party libraries (and getting them fixed for the new version), and offering to continue supporting those 4 packages on all 4 platforms as incompatibilities arise (eg, Linux libraries change, Apple 10.9 adds new signature requirements, etc) would be much more compelling.
But that is a LOT of work. From what I've seen over the last several years, the odds of anyone stepping up to actually do that work are very slim.
However, fixing this specific bug is not very much work. I plan to do it.... assuming Bill responds with specific info, which I'm sure he will.