Notice of compiling egcs-1.0.1

Notice of compiling egcs-1.0.1

[FireBird (Richard Rose)]

I did it! After only a few 4 hour compiling sessions, I finally did it.

The problem was, I had explicit .'s in my environment variables. I know
that it says in the documents online that it won't compile if you have
them, but the way I, and no doubt lots of other Linux users, got egcs was
by ftp, so I never saw the documents. In my opinion, a great improvement
for the make process, is to get make to call a program which scans the
environment strings for explicit .'s and tells the user. It would have
saved me 12.5 hours.

The documentation said that you wanted the output from config.guess, so
here it is: i586-pc-linux-gnulibc1
This guess is correct.

If you are wondering why you have recieved two copies of this e-mail, it
is because my ISP (Zen Internet, in the UK) have a habit of sending
SIGHUP's at random times, and I was halfway through sending this e-mail
the first time when I reieved one, and I was not sure whether it had gone,
so I sent it again...

Keep up the good work..
firebird@scitech.co.uk

Re: Notice of compiling egcs-1.0.1

[Craig Burley]

>The problem was, I had explicit .'s in my environment variables. I know
>that it says in the documents online that it won't compile if you have
>them, but the way I, and no doubt lots of other Linux users, got egcs was
>by ftp, so I never saw the documents. In my opinion, a great improvement
>for the make process, is to get make to call a program which scans the
>environment strings for explicit .'s and tells the user. It would have
>saved me 12.5 hours.

Whoah, that's incredible, another example of "if you know something
that a user does to operate your product won't work at some point,
try to make you product fail in that case as soon, and clearly,
as possible".

For my own part, I have *no* idea what "explicit .'s in my environment
variables" means, as I, a Linux user (Intel and Alpha), also get my
egcs copies via ftp.

Could someone who understands this issue please implement some
early "trap" in the build process (and, if the problem can make
configuring misbehave, early on there as well) so more users
don't run into this problem?

        tq vm, (burley)

Re: Notice of compiling egcs-1.0.1

[Jeffrey A Law]

  In message < 199805051631.MAA09898@melange.gnu.org >you write:
  > Whoah, that's incredible, another example of "if you know something
  > that a user does to operate your product won't work at some point,
  > try to make you product fail in that case as soon, and clearly,
  > as possible".
Yup.  Nailing this down took quite a bit of work on Jim's part
several months ago.  Believe me, it had lots of people stumped.

  > For my own part, I have *no* idea what "explicit .'s in my environment
  > variables" means, as I, a Linux user (Intel and Alpha), also get my
  > egcs copies via ftp.
setenv LIBRARY_PATH x:.

Gets you an explicit one after "x" :-)

setenv LIBRARY_PATH :x:

Gets you two implicit ones (one before "x", one after "x").



  > Could someone who understands this issue please implement some
  > early "trap" in the build process (and, if the problem can make
  > configuring misbehave, early on there as well) so more users
  > don't run into this problem?
It's a "make" time issue, not a configure time issue.  As often as
it's come up I don't believe trying to work around users with losing
environments in the makefiles is the right way to approach the problem.

Unfortunately, we can't force folks to read the FAQ (which is
available both on the web *and* in the distributions)....

jeff

Re: Notice of compiling egcs-1.0.1

[Joern Rennecke]

> Unfortunately, we can't force folks to read the FAQ (which is
> available both on the web *and* in the distributions)....

Well, at least for the web interface, we could require them to finish a
test on the FAQ before they come to the download page ;-)

Re: Notice of compiling egcs-1.0.1

[Craig Burley]

>It's a "make" time issue, not a configure time issue.  As often as
>it's come up I don't believe trying to work around users with losing
>environments in the makefiles is the right way to approach the problem.
>
>Unfortunately, we can't force folks to read the FAQ (which is
>available both on the web *and* in the distributions)....

Huh, well, just so you know, I couldn't find this *anywhere* in the
INSTALL subdirectory for egcs-19980425.  Searching for strings
like "setenv", "environment", "path", and "LD_LIBRARY_PATH" turned
up nothing that I would see as telling me how I'd screwed up my
environment, were I to run into this problem.

If I knew more about the problem, I'd be willing to write an early
test-with-diagnostic for it myself and submit that as a patch to
egcs.  I'm sure it'll take less of my time than of the next person
to run into this problem without such a patch present; even less
of somebody else's time, if that person knows the issues and such
better than I do, but at least I'm volunteering.  (If I'm finally
pointed to a good explanation of the problem, that might be all
I need to quickly write a decent test for it.)

Please remember: *users* with losing *environments* are often not
the same people who set up those environments in the first place!

(Not sure what is meant by "losing environments in the makefiles",
since I thought things like LD_LIBRARY_PATH were part of the shell,
or process, environment.)

        tq vm, (burley)

Re: Notice of compiling egcs-1.0.1

[Gerald Pfeifer]

On Wed, 6 May 1998, Craig Burley wrote:
> Huh, well, just so you know, I couldn't find this *anywhere* in the
> INSTALL subdirectory for egcs-19980425.  

Problem might be the "Last modified:  December 2, 1997" in faq.html. :-(


For now: Could someone with CVS access copy over the version of faq.html
from wwwdocs/INSTALL into egcs/INSTALL? 

In general, egcs/INSTALL is out of sync with wwwdocs/INSTALL. Shouldn't
the former be an exact mirror of the latter?

Gerald
-- 
Gerald Pfeifer (Jerry)      Vienna University of Technology
pfeifer@dbai.tuwien.ac.at   http://www.dbai.tuwien.ac.at/~pfeifer/

Re: Notice of compiling egcs-1.0.1

[Jeffrey A Law]

  In message < 199805061429.KAA22340@melange.gnu.org >you write:
  > Huh, well, just so you know, I couldn't find this *anywhere* in the
  > INSTALL subdirectory for egcs-19980425.  Searching for strings
  > like "setenv", "environment", "path", and "LD_LIBRARY_PATH" turned
  > up nothing that I would see as telling me how I'd screwed up my
  > environment, were I to run into this problem.
Thanks.  I really thought that one was in the FAQ distributed with
egcs-1.0.{1,2}.  Apparently not.

The good news is I just tweaked the release script so that it
gets copies of the most recent faq & install instructions to
prevent this from happening again.

  > Please remember: *users* with losing *environments* are often not
  > the same people who set up those environments in the first place!
Often true.

  > (Not sure what is meant by "losing environments in the makefiles",
  > since I thought things like LD_LIBRARY_PATH were part of the shell,
  > or process, environment.)
Right, but environments are inherited by children.

ie, if I have GCC_EXEC_PREFIX in my environment, then I invoke make,
it gets passed down to the children of make.

jeff