[avr-gcc-list] Re: AVR toolchain patch hell

[Denis Chertykov]

Eric Weddington <address@hidden> writes:

> Hello All,
> 
> I've sent this to everyone individually because I'm unsure who is subscribed
> to which avr mailing list (avr-libc-dev, avr-gcc-list, CCed).

I'm subscribed only to address@hidden
I'm even not subscribed to address@hidden

> We're starting to be in AVR toolchain patch hell, and especially for GCC
> 4.1.x. Here are some points to consider:
> 
> - Joerg has a set of patches that he keeps for the FreeBSD AVR toolchain in
> the Ports system.
> - I have a set of patches for the WinAVR distribution. Joerg and I at least
> try to do some level of coordination.
> - Bernd and Rolf have another set of patches for the AVR-Ada project:
> <http://svn.sourceforge.net/viewvc/avr-ada/trunk/patches/>. Some are avr-ada
> specific, some patches are for AVR GCC bug reports. I know that Bernd and
> Rolf help out sometimes on the base AVR toolchain, and they do some level of
> coordination on patches. And Bernd is copied on all my traffic on GCC
> Bugzilla.

I can commit any patch from the person with signed GNU paper.

> - Bjoern has other patches that are in process
> - There are currently 56 GCC bugs for the AVR target

I know about them. As I remember, only two of them are really critical.
They are:
PR 19636 - probably it's a reload bug not an AVR port;
bug with use of libssp (you need to pass --disable-libssp).

> - There are currently 3 Binutils bugs for the AVR target

I don't know about  :(

> - There are currently 11 avr-libc bugs
> - There are currently 2 avr-ada bugs
> - There are currently 3 avrdude bugs
> - Denis and Marek are the sole maintainers for the AVR target, and Marek has
> said recently that he doesn't have time for any more AVR toolchain
> stuff.

Anatoly Sokolov can be the maintainer.

> - Joerg has been doing a lot of maintenance on avr-libc, avrdude, and
> avarice.
> - We only see an occasional Linux maintainer (gentoo, debian) and they
> usually just grab a limited number of existing patches and don't seem to
> help much upstream.
> - We have no GDB maintainer since Ted Roth left, and the avr port of GDB is
> starting to bit-rot.

It's very bad. I'm worry about GDB port.

> That leaves one person (Denis) as the bottleneck for committing patches into
> binutils and gcc for the avr target. I note that Bjoern has been able to get
> some of the other gcc people to do commits.

Generally, they (GCC peoples) don't bothered about AVR port. It's just
an 8-bits microcontroller. They are right. 

> I'm not trying to place blame on anyone; I just want to describe the current
> situation. I know that every one on this list has been extremely valuable in
> helping to move the avr toolchain forward, and for that, I am extremely
> grateful.
> 
> But right now, it is difficult to take a stock released version of the tools
> and build a decent toolchain from them. Everything has to be patched, or
> really should be patched. Some of that can be blamed on Atmel Norway for
> releasing new versions of chips every quarter. But there isn't much
> coordination about patches, and it's extremely difficult to track down all
> available patches, get them working for release X, and especially getting
> them rolled into the mainlines of the various projects. I'd also finally
> like to get avr-ada into the mainstream of all the distributions; Bernd,
> Rolf, I know I've said this before.
> 
> So I am requesting of you: what can we do to help this situation?

I can review/comment/reject/apply patches to GCC mainstream.

> - Can we get more people to be official maintainers of the avr port of
> binutils and gcc?

I think that I can change Marek with somebody else.

> What else?
> 
> I'm also committing more time now to the overall toolchain. I want to help
> the overall situation, especially with all the various bug reports

I can help you with bug reports.

I have the one big problem. The AVR port havn't working GCC testsuit.

Denis.