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

Re: gcc?

Nikita Schmidt <cetus@snowball.ucd.ie> writes:

> I tried egcs and found it very nice.  It does not need any pesky patches
> and is much more aware of linux-alpha (which the FSF gcc does not even

Cool.  I was wondering if anyone had tried it for Debian yet.

Oh, and by way of introduction:

     My name is Ken Raeburn.  I work for Cygnus Solutions, in the
     compiler group at the moment (and on binutils and kerbnet in
     the past), but I do not work on EGCS.

     I've just joined the list a little while ago, having
     recently bought an Alpha from a friend who ran out of time
     for NetBSD/Alpha hacking.  I'm running RedHat on it at the
     moment, since neither NetBSD nor Debian had an install setup
     I could get to work at the time.  I use both NetBSD and
     Debian Linux on other computers I own.

     I'm still hoping to move to Debian on my Alpha, but
     unfortunately I can't contribute lots of time right now for
     debugging random packages.  So I'm sorta waiting for you
     guys to do it right now, sorry.  When I get a few other
     projects finished and off my to-do list, I'd like to help

> support).  Although egcs is considered development version and does not
> have public releases, I believe that linux-alpha support in it is way
> more stable than the one in patched-gcc.

Their plan is to have releases in addition to the weekly snapshots.
They just haven't got to release #1 yet.  I don't know if the current
target release date has been advertised publicly (or if it's been
changed since last I heard), so I probably shouldn't say more.

>From what I've seen of gcc development, on gcc2 and egcs (which I
don't have time to follow very actively at the moment), I'm not
surprised at this situation (solid linux-alpha support in egcs when
gcc2 has none).  The gcc2 situation should improve after 2.8 comes
out, but that remains to be seen.

> Anyway, at the moment it is better to use gcc (I can't compile libg++
> with egcs, for example), but later it may be worth considering egcs as our
> compiler of choice.

Some g++ support libraries are included with egcs now, but I think
libg++ is mostly considered deprecated.  There may be a few things it
has that the other libraries don't, but last I heard, no one had
expressed enough interest to pick up their maintenance.  (But I'm
*not* up to date on egcs mail, it may have happened.)  So I doubt
you'll see a libg++ fixed to work with egcs.

Also note that the recent versions of binutils (snapshots definitely,
last release maybe) will attempt to treat an input file that doesn't
have a recognizeable object or archive format as a linker command
script.  The main point of this was that you could include a "library"
which was really a script telling the linker to bring in N other
libraries.  So if the libraries with egcs are adequate, you could
define a fake libg++ library which tells the linker to pull in all the
egcs libraries.  That may make porting from gcc2/libg++ to egcs

There may be a minor problem with overlap between egcs and glibc
releases.  I think both provide versions of a "libio" library.

I could be wrong about some of this...is anyone here actively
following the egcs lists?

What other problems have you seen with egcs?  (I.e., what else would
be needed for you to be able to consider making the switch soon?)  Get
them all reported (directly to egcs folks, not to me -- I just ask out
of curiosity), and maybe the first release will be good enough to use
for debian-alpha until the gcc2 crowd gets a release out with working
alpha-linux support.

-- Ken
   (not speaking for my employer, Cygnus Solutions, who's spearheading
    the EGCS work, which I don't have time to follow, nor speaking for
    the EGCS group, just speaking as one lone individual who has
    occasionally paid attention to what's going on there)

TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to
debian-alpha-request@lists.debian.org . 
Trouble?  e-mail to templin@bucknell.edu .

Reply to: