Re: egcs & gcc 2.8 ??

Re: egcs & gcc 2.8 ??

[Yotam Medini]

>     >  http://www.cygnus.com/egcs/egcs-1.0.html:
>     > 
>     >    egcs-1.0 also contains many improvements and features not found in
>     >    gcc-2.7 and even the soon to be released gcc-2.8 compilers.
>     > 
>     > Could someone please educate me about the relations between 
>     > egcs and gcc 2.8?
>   It's pretty simple.
>
>   egcs-1.0 is based on an early August snapshot of the gcc-2.8 development
>   sources with many of our own enhancements & bugfixes.
>
>   The current snapshots are based on the soon to be release gcc-2.8 (again
>   with our enhancements and bugfixes).
>
>   jeff

Many of egcs users, I believe, assume that a code 
compiled with egcs's gcc/g++ could be compiled eventually
with the official gcc-<l>.<m>.<n>.
Therefore, we would all like to help with early feedback on further 
gcc development.

So the questions are really:

  * How gcc's official release is determined.
  * What are the egcs features that will not make it
    into the next gcc release.
  * What is the expected time length for a bug-fix in some egcs snap-shot
    to propogate to the official gcc release.

Much thanks -- yotam

Re: egcs & gcc 2.8 ??

[Jeffrey A Law]

  In message < 199712111819.KAA02784@telaviv.tmai.com >you write:
  > Many of egcs users, I believe, assume that a code 
  > compiled with egcs's gcc/g++ could be compiled eventually
  > with the official gcc-<l>.<m>.<n>.
  > Therefore, we would all like to help with early feedback on further 
  > gcc development.
That's probably not a safe assumption.  We're moving forward at a much faster
pace than gcc2.  Odds are we've already got some g++ features in egcs that
are not in gcc2.


  > So the questions are really:
  > 
  >   * How gcc's official release is determined.
I don't know how to answer that.  That's one of the problems with the whole
gcc-2.8 process, nobody knows how/when releases will happen.  I (and othes)
complained more than once about this problem, but nothing ever happened.


  >   * What are the egcs features that will not make it
  >     into the next gcc release.
That depends mostly on the gcc2 folks.  It's mostly out of our hands.

  >   * What is the expected time length for a bug-fix in some egcs snap-shot
  >     to propogate to the official gcc release.
Again, depends on the gcc2 folks.

jeff

Re: egcs & gcc 2.8 ??

[Joe Buck]

>   In message < 199712111819.KAA02784@telaviv.tmai.com >you write:
>   > Many of egcs users, I believe, assume that a code 
>   > compiled with egcs's gcc/g++ could be compiled eventually
>   > with the official gcc-<l>.<m>.<n>.
>   > Therefore, we would all like to help with early feedback on further 
>   > gcc development.
> That's probably not a safe assumption.  We're moving forward at a much faster
> pace than gcc2.  Odds are we've already got some g++ features in egcs that
> are not in gcc2.

No, I think it is a safe assumption (that C++ code that compiles with egcs
g++ will compile eventually with gcc2) -- the intent is that the C++ front
end be installed in both egcs and gcc2, as I understand things.  At any
given point in time, egcs may be slightly ahead.  *Warning*: this assumes
that you attempt to write standard C++, not just whatever the compiler
will accept; in the past, g++ has accepted invalid input due to bugs and
folks complain when the next release fixes the bug and rejects the invalid
input.

>   > So the questions are really:
>   > 
>   >   * How gcc's official release is determined.
> I don't know how to answer that.  That's one of the problems with the whole
> gcc-2.8 process, nobody knows how/when releases will happen.  I (and othes)
> complained more than once about this problem, but nothing ever happened.

Richard Kenner decides -- it seems he now wants to get a release out ASAP,
but there are still a couple of serious bugs.  (and thanks to the recent
merge these bugs are now in the latest egcs snapshot).

>   >   * What are the egcs features that will not make it
>   >     into the next gcc release.
> That depends mostly on the gcc2 folks.  It's mostly out of our hands.

haifa and -frepo, to name two.  Otherwise they aren't going to be all that
different, I think.

>   >   * What is the expected time length for a bug-fix in some egcs snap-shot
>   >     to propogate to the official gcc release.
> Again, depends on the gcc2 folks.

I suspect that many egcs-generated bug-fixes will appear relatively
quickly in the gcc2 *snapshots*.  How long it will take before a release
occurs is another matter.

Re: egcs & gcc 2.8 ??

[Jeffrey A Law]

  In message < 199712112351.PAA23325@atrus.synopsys.com >you write:
  > No, I think it is a safe assumption (that C++ code that compiles with egcs
  > g++ will compile eventually with gcc2) -- the intent is that the C++ front
  > end be installed in both egcs and gcc2, as I understand things.  At any
  > given point in time, egcs may be slightly ahead.  *Warning*: this assumes
  > that you attempt to write standard C++, not just whatever the compiler
  > will accept; in the past, g++ has accepted invalid input due to bugs and
  > folks complain when the next release fixes the bug and rejects the invalid
  > input.
That's actually a more accurate statement than mine.  Thanks Joe.

  > > That depends mostly on the gcc2 folks.  It's mostly out of our hands.
  > 
  > haifa and -frepo, to name two.  Otherwise they aren't going to be all that
  > different, I think.
I think the EH code in egcs is further along, including more accurate flow
graphs, which are pretty important for both code correctness and valid warnings.
We'll have global cse & partial redundancy elimination soon, we've also got
several improvements to the loop optimizer.


  > I suspect that many egcs-generated bug-fixes will appear relatively
  > quickly in the gcc2 *snapshots*.  How long it will take before a release
  > occurs is another matter.
Quite true.

Thanks for helping clarify things Joe.  Sometimes I'm a little too terse.

jeff