🚀 go-pugleaf

If someone can find which line of Elm is miscompiled (or at least
localize it to a few lines), and send me a gcc bug report with a test
case in the form of CPP output, I can fix this bug.

Otherwise, there is nothing I can do.

The first step in trying to localize this is to try compiling one Elm
source file at a time with the old version, and testing the result,
until it starts to work.  Use the option -fpcc-struct-return when
compiling with the old GCC.

The last file recompiled with the old GCC is the one where the problem
is.  Or at least, one problem.

Then, try debugging what happens in that source file.  You can assume
that all functions in other files execute correctly, so you don't need
to bother really looking at them.  Just step through them and assume
they do what they ought to.

If you have a suspicion about a particular function, and the function
is not too large, report that to me right away.  That *may* be enough
to enable me to fix the problem.  Then continue trying to find the
miscompilation, since perhaps just the function name is not enough.

For example, if you can determine that the error happens within a
particular function for substituting into a template string, send me
CPP output that I can compile directly, including that function.

Please, everyone, try to work on this today if you can, since that
will enable us to put a reliable GCC on our CD-ROM.


I'm getting really odd behaviour compiling Elm 2.4[PL21] on our system
(Sun Sparc, SunOs 4.1.2) since we went to gcc 2.4.1 from gcc 2.3.3

Under 2.3.3 we had no problems with Elm, but since we got 2.4.1, Elm
compiles fine, with no errors, but on running, the header line at the top
of the index screen is corrupt and looks like this :

 is '%s is '%s' with %d messages [ELM %s]' with 344328 messages [ELM
/tsunb/home
/mail/samuele]