Linaro GCC plans Meeting 2011-03-02 10:00 - 11:00 EET

Attendees:
	Alexander Sack      (Linaro)
	Michael Hope        (Linaro)
	Jan-Simon Moeller   (Linux Foundation)
	Al Nikolov          (Nokia)
	Alexander Kanevskiy (Nokia)
	Carsten Munk        (Nokia)
	Jarmo Kant          (Nokia)
	Juha Kallioinen     (Nokia) - Chair, Secretary
	Leonid Moiseichuk   (Nokia)
	Martti Lumme        (Nokia)
	Riku Voipio         (Nokia)

Agenda:
	* Introductions
	* Linaro toolchain plans and MeeGo 1.3 synchronization
	* AOB

= General about upcoming work and plans for Linaro and MeeGo =

Linaro makes verified toolchain releases monthly. The current work is
based on the current FSF gcc 4.5.2 release + changes for 4.5.3 in the
subversion branch.

MeeGo 1.3 is planning to take gcc 4.5.2 or 4.5.3 into use. MeeGo 1.3
work starts at the end of March, but the toolchain can be updated up to
a yet-unset deadline, which might be at the end of June/July 2011.

On MeeGo side we plan to evaluate the Linaro toolchains hopefully
monthly soon after each new release is available. How the evaluation
will be executed is to be planned. In order to help the project we
should report new evaluation findings after a quick triaging in
launchpad https://bugs.launchpad.net/gcc-linaro/

The current FSF gcc work is focusing on getting the first 4.6 release
out, which may cause delays in the release of gcc 4.5.3. A new
microversion release usually comes out in 3 - 6 months after the
previous release. Gcc 4.5.2 was released in Dec/2010, so there's a
chance 4.5.3 comes out in time for MeeGo 1.3.

Once gcc 4.5.3 is out, Linaro will continue working on improving
that. Michael notes that there has been a great demand for a good gcc
4.5-based ARM toolchain and Linaro is focused on delivering that.

Once gcc 4.6 is out, Linaro will start working on that as the
development trunk, but will still keep improving the 4.5 in a
maintenance branch. The exact plan about how to divide work between
the 4.5 and 4.6 versions is still a bit unclear.

As a general note it was mentioned that the first release version FSF
gcc 4.6.0 should not probably be used in a production environment. It
will be good for evaluation purposes, though.

= Miscellaneous topics =

* Linaro does not have a recommended combination of
  eglibc/binutils/gcc versions. Recent versions of (e)glibc and
  binutils are known to work just fine.

* A question about eglibc vs glibc. Michael tells that he/Linaro
  prefers eglibc since the project is more friendly towards accepting
  ARM changes. It also offers more options to leave features out,
  which might be useful when looking to save space or not wanting to
  worry about possible security issues with some features.

  Glibc + glibc-ports offers the same functionality without all the
  configurability options. MeeGo has chosen to stick with glibc.

* Linaro's efforts are focused in delivering a good ARM toolchain, but
  they also verify that there is no breakage or performance
  degradation in x86/x86_64 code because of Linaro's changes.

* Since Linaro does not make any performance improvements for x86,
  they could possibly accept contributions also to the x86 side of the
  toolchain if someone were to submit them. This topic was just
  quickly thrown out there and will require approval from Linaro
  management. These contributions would not be maintained by Linaro,
  but by the submitter. For MeeGo it would be great to be able to use
  the same toolchain for x86 and ARM.

* Linaro also delivers ARM specific improvements to GDB and
  qemu. MeeGo will definitely want to look at the GDB improvements.

* About the versions of gcc's build-dependencies: The FSF
  documentation lists the minimum version requirements for the
  build-deps and those or newer should be fine. When built using the
  three-stage bootstrap method, gcc runs so many internal tests to
  verify its correctness that any problems with the build-deps should
  be caught already at that point.

Cheers,
 Juha Kallioinen