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

Re: Anybody working on xfree 3.3.2??


>> troff never worked?? Or did someone introduce a feature that
>> confuses gcc
>For the longest time (certainly as long as I've been building stuff
>for m68k) troff has been broken in that it would spinlock in obscure
>cases, notably while building elm and nethack or formatting some man
>pages (super's IIRC).  I tracked it down to gcc misoptimizing code so
>that the return value from a function got lost.  As a result, the last
>couple of versions of groff have been compiled without optimization on
>m68k.  I reported it to Andreas and he said he couldn't reproduce the
>problem (which is why I think egcs/2.8.1 will fix it).
We'll see that, eventuall :-) But if Andreas can't reproduce it, it won't 
happen anymore. 
Having to build troff without optimization is tragic (explains the sloooow
manpage formatting) but it might have been sufficient to only build that 
one function non-optimized. Whatever, I suppose we can't back out of the 
egcs deal anymore.

>nsgmls was totally hosed, it would do nothing but print a `C' and
>exit, causing no ends of fun for any SGML based documentation.

Ok, if a new compiler fixed it, fine...

>> I'm still wondering why egcs is used, not gcc 2.8. Whatever.
>Well at the time Galen made the choice 2.8* didn't exist and even when
>it did certainly 2.8 suffered from some pretty bad problems (though
>2.8.1 is meant to be a _lot_ better), also there is a lot of

Didn't know it's been that long that these details have been hashed out...

>`cool'-factor behind egcs because it follows the much vaunted bazaar
>method as opposed to the Evil/RMS-style (TM) cathedral method of
>development, which may have, IMHO, unduly influenced people's lobbying
>for egcs.
Hilarious. May I propose another smart move to boost the 'cool' factor? 
We should do everything in C++ or Java in future. Including the kernel,
of course.

I thought egcs and gcc 2.8 are related the same way as 2.1 and 2.0 Linux
kernels - one is experimental, the other stable (well, 2.8.0 was a disaster
from what I've heard, but so were some 2.0 kernels). The bazaar vs. cathedral
thing is rather a religious issue. I was very amused to hear that it had 
influenced Netscape, but I'm less thrilled if the compiler actually proves 
to be the problem. Which isn't exactly the case - how's the work on that
case going??

Maybe Stefan should post his experience to linux-m68k, or post a working and a 
broken kernel on some FTP site for a detailed analysis?


To UNSUBSCRIBE, email to debian-68k-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

Reply to: