librelist archives

« back to archive

strange behaviours

strange behaviours

From:
Cyril Adrian
Date:
2013-03-18 @ 15:49
 OK, I could reproduce the strange behaviours we see sometimes (strange
error messages, seemed to happen only with -O2 or higher gcc optimizations)

I bissected gcc's O2 flags and found the culprit: -fgcse

The installation works fine with -O2 -fno-gcse

It is another clue pointing to the actual compiler problem, although I
still don't know what it really is.

Meanwhile I'm going to commit the change in install.sh to see how ET fares.

Cheers,





*Cyril ADRIAN** (from office)*
 [image: Google Plus] <http://plus.google.com/100388810006463519079> My
latest G+: 
http://computerhistory....<http://plus.google.com/100388810006463519079/posts/DkqWMSU7bVj>
My G+ <http://plus.google.com/100388810006463519079> -
Posts<http://plus.google.com/100388810006463519079/posts/>- Add
to Circles <http://plus.google.com/100388810006463519079/about/> - 17:00
Feb-26 <http://plus.google.com/100388810006463519079>
  Get this email app!

<http://www.wisestamp.com/apps/plus?utm_source=extension&utm_medium=email&utm_term=plus&utm_campaign=apps>

  <http://fr.linkedin.com/in/cadrian/>
  Get a signature like this.

<http://r1.wisestamp.com/r/landing?promo=35&dest=http%3A%2F%2Fwww.wisestamp.com%2Femail-install%3Futm_source%3Dextension%26utm_medium%3Demail%26utm_campaign%3Dpromo_35>
CLICK

HERE.<http://r1.wisestamp.com/r/landing?promo=35&dest=http%3A%2F%2Fwww.wisestamp.com%2Femail-install%3Futm_source%3Dextension%26utm_medium%3Demail%26utm_campaign%3Dpromo_35>
​

   <http://twitter.com/cadbart>  Latest tweet:
Follow @cadbart <http://twitter.com/cadbart>    • https://t.co/m0kegfkOHW
fix compile_to_c.new removal, only when
needed<http://twitter.com/cadbart/statuses/313290060197928960>
• https://t.co/ObfA8U8xV1 changed a few dubious
sizeofs<http://twitter.com/cadbart/statuses/313287796817268736>
• https://t.co/eMyb7PLU29 Merge branch 'master' of
http://t.co/IcRBj5g3xu:cadrian/Liberty Conflicts:
install.sh<http://twitter.com/cadbart/statuses/313269453641101312>
View
more <http://twitter.com/cadbart>
   Get this email app!

<http://www.wisestamp.com/apps/twitter?utm_source=extension&utm_medium=email&utm_term=twitter&utm_campaign=apps>

Re: [libertyeiffel] strange behaviours

From:
Raphael Mack
Date:
2013-03-20 @ 20:13
Hi,

Am Montag, den 18.03.2013, 16:49 +0100 schrieb Cyril ADRIAN: 
> OK, I could reproduce the strange behaviours we see sometimes (strange
> error messages, seemed to happen only with -O2 or higher gcc
> optimizations)
> 

> I bissected gcc's O2 flags and found the culprit: -fgcse

Interesting. I did not expect common subexpression elimination to be
responsible for these kind of effects. But of cause it is not clear how
aggressive gcc does it. It feels as if our C code relies on some side
effects during expression evaluation - but in the GC? 
> The installation works fine with -O2 -fno-gcse

Also for me, no problem on both machines.
> 
> It is another clue pointing to the actual compiler problem, although I
> still don't know what it really is.

The GC does not rewrite pointers, does it?
> 
> Meanwhile I'm going to commit the change in install.sh to see how ET
> fares.

Well done!

Regards,
Rapha 

Re: [libertyeiffel] strange behaviours

From:
H. Zwakenberg | Ocean Consulting GmbH
Date:
2013-03-20 @ 20:24
 Hi,

 same issues persist with no GC activated?  Is it really the GC code?  
 If so, why not use the Boehm code, which is well proven?

 Hans



 Am Mittwoch, den 20.03.2013, 21:13 +0100 schrieb Raphael Mack 
 <ramack@raphael-mack.de>:
> Hi,
>
> Am Montag, den 18.03.2013, 16:49 +0100 schrieb Cyril ADRIAN:
>> OK, I could reproduce the strange behaviours we see sometimes 
>> (strange
>> error messages, seemed to happen only with -O2 or higher gcc
>> optimizations)
>>
>
>> I bissected gcc's O2 flags and found the culprit: -fgcse
>
> Interesting. I did not expect common subexpression elimination to be
> responsible for these kind of effects. But of cause it is not clear 
> how
> aggressive gcc does it. It feels as if our C code relies on some side
> effects during expression evaluation - but in the GC?
>> The installation works fine with -O2 -fno-gcse
>
> Also for me, no problem on both machines.
>>
>> It is another clue pointing to the actual compiler problem, although 
>> I
>> still don't know what it really is.
>
> The GC does not rewrite pointers, does it?
>>
>> Meanwhile I'm going to commit the change in install.sh to see how ET
>> fares.
>
> Well done!
>
> Regards,
> Rapha