AW: [avr-gcc-list] Don't use gcc 3.4.4, use 4.0.1

[Haase Bjoern (PT-BEU/EMT) *]

Hi,

My present opinion is that #21990 is a hidden problem in reload.c (i.e. the 
worst of all the source files within the gcc tree, accessible only to initiated 
gurus). I have succeeded in reproducing the bug only in conjunction with 
strange register constraints in inline asm statements. Mainly, it seems that 
one should avoid two simultaneous in/out operands with the "+" modifier (e.g. 
"+x").

We are using 4.0.1 sucessfully for our applications right now without observing 
any serious problem. I, however, still feel kind of uneasy with 4.x.x . I'd 
suggest: Try it but be careful.

Yours,

Björn

-----Ursprüngliche Nachricht-----
Von: address@hidden [mailto:address@hidden Im Auftrag von E. Weddington
Gesendet: Freitag, 29. Juli 2005 01:59
An: Andy Warner
Cc: address@hidden; Björn Haase; address@hidden; Denis Chertykov
Betreff: Re: [avr-gcc-list] Don't use gcc 3.4.4, use 4.0.1

Andy Warner wrote:

>E. Weddington wrote:
>  
>
>>[...]
>>I'd like some other issues in GCC 4.x to be cleaned up for the AVR port 
>>before including it in WinAVR, especially DWARF2 issues. Here is a list 
>>of known AVR GCC bugs:
>><http://rtems.org/phpwiki/index.php/GCCAVRBugs>
>>    
>>
>
>As a Linux-hosted user, is 4.0.1 stable enough for me to spend
>some time getting to know it ? I'm running 3.4.3 right now, and
>am very happy with it - but smaller/faster code is always welcome.
>I'm cool building it all from source.
>  
>

Personally I would still hesitate to use it in any type of production 
environment because of these two GCC bugs:
<http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21990>
<http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21107>

Bug #21990 is a "wrong code" type bug and was submitted by Björn Haase. 
He stated in the beginning:
"I have observed a wrong code bug that I judge to be so serious that 
IMHO one should discourage use of the avr port for 4.x.x until it is 
resolved."

Bug #21107 is an "ICE on valid" bug submitted by Martin Kögler and is a 
regression for 4.0/4.1.

Neither of these two bugs have been fixed yet.

IMHO. YMMV.

Eric Weddington


_______________________________________________
AVR-GCC-list mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/avr-gcc-list