[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[avr-gcc-list] Re: AVR bug fix status

From: Andrew Hutchinson
Subject: [avr-gcc-list] Re: AVR bug fix status
Date: Sun, 16 Mar 2008 13:56:36 -0400
User-agent: Thunderbird (Windows/20080213)

I don't have any problem. I posted list so as not to forget which I had fixed and posted

But I cant fix much more unless HEAD catches up. Some of the bug are same - I found some duplicates but suspect there are more.

I have more fixs than I have posted in the email - but these need other problems fixed before a patch would make sense.

I posted response to GCC maintainer comment on same day. I will chase them again.

That bug was regression from 4.2 - but I cannot change its status.

Also many bugs are marked UNCONFIRMED - I cannot change status.

Having bug CONFIRMED and REGRESSION gets more attention.

Beware, Combine bug is nasty - if you are lucky, it will cause SEG fault - if not bad code - no warning.

We could do with a volunteer to create AVR testcase directory in GCC - so we could add testcases.
It makes it easier to find potential duplicates.


Weddington, Eric wrote:
Hi Andy,

I understand your reluctance against 4.3.0. My problem is that I need to
have support for the new xmega devices. All of that support has been
written for 4.3.0, so I will be releasing the next WinAVR with 4.3.0.

I have your patch for bugs #19636,24894,31644,31786: spill failure ICE.
I have your patch for bugs #27192,35013: call through function pointer.
I am doing a test build now with your patch for bug #32871.
The patch for bug #34412 has been committed.

For bugs #34916,35519, IIRC there was a GCC maintainer that wasn't happy
about your proposed patch. So you'll either have to convince them that
it is ok, or rework your patch.

For the other bugs (27663,35542; 35507), these are missed optimizations.
They are important, but I would rather focus on getting your current
fixes committed as those are more important bugs to get fixed.

I do not have write permissions in GCC (I don't yet have the FSF
Copyright Assignment paperwork done), so I can't do a whole lot other
than use your patches in WinAVR. Anatoly is one of the AVR port
maintainers, so he will have to get these committed. I know that both
Anatoly and I have a number of patches in our backlog to get committed,
so I hope that we can get to your patches soon. I know that Anatoly is
working through his backlog. For example the mega256x patch was
committed to GCC recently (for 4.4). All I can ask of you is to have
patience. But please keep bugging us! :-) I definitely want to see that
these bugs get fixed.

Eric Weddington

-----Original Message-----
From: Andrew Hutchinson [mailto:address@hidden Sent: Sunday, March 16, 2008 10:06 AM
To: address@hidden; Weddington, Eric; Anatoly Sokolov
Subject: AVR bug fix status

Here is status of the bugs I have looked at.

Bug status needs updating to status in Bugzilla

For example, confirm and update known to work/known to fail information. I cannot do this.

Some of the bugs are serious and generally I would advise against using 4.3 - especially with long operands.


[avr] gcc is pushing too many registers

Patch created and submitted  with Anatoly since 3/2/2008

Various BASE_POINTER spill failure ICE  bugs


Patch created and submitted  with Anatoly since 1/16/2008
ICE in extract_insn, at recog.c:1990 AVR Tiny Stack


fixed and committed
Incomplete check for pm() annotation/call through function pointer goes to wrong address


Patched posted 2/16/2008

The following GCC bugs are not due to AVR backend - but cause problems!

[Bug middle-end/35519] COMBINE repeating same matches and can SEG fault]
gcc.c-torture/execute/pr27364.c fails with -O1, -O2 and -Os


Combine bug posted with patch 3/9/2008
fwprop only propagates one operand
missed-optimization transforming a byte array to unsigned long (and perhaps many others)


Bug reported 3/11/2008 with suggested patch

[avr] 4.3.0: size of small function increases from 2 to 29 words

Outline solution posted 3/9/2008 - needs patch.

reply via email to

[Prev in Thread] Current Thread [Next in Thread]