On 3/30/07, Mike Stump <[EMAIL PROTECTED]> wrote:
On Mar 30, 2007, at 12:32 PM, Null Heart wrote:
> I was just poking around with the latest snapshot for fun
Two thoughts come to mind. First, qualify your system with a known
to build, known to be good compiler. Build it 20 times, if it never
fails to build, you probably have a good system. If it ever fails,
toss the system out and get a real machine. :-)
My video card cost more than most machines, and I've been there and done that.
> when I came across a huge problem: the make would fail without reason.
If you need help reading a build log, well, this isn't quite the
right place for it. Look for '***' in the log as a very rough
approximation. Does sound like you did track it down to the right
failed file.
... No file failed. GCJ did not give an error. The only error was the
generic make error. I know, I repeated the process and varieties for
hours trying to track down an error on my part. I get "make[3]: ***
[gnu/javax/swing/text/html/parser/HTML_401F.lo] Error 1" and no other
error. I do not even get the memory error that others got.
> Running the included libtool command manually without the "-O2"
> worked perfectly.
Run the command with the -O2 option, does it fail? If so, does it
always fail if you run it 5 times? If so, that bug is probably a run
of the mill compiler bug. Feel free to file a bug report. Do make
sure you have enough memory and paging file space on the system though.
The "-O2" is there, apparently, by default. It took removing it to
finish the compile. The bug may be in the compiler, but I don't have
enough specific information to file a bug report or I already would
have. I have two gigabytes of memory, I think I'm good.
Also, you should be able to check the gcc-testresults mailing list, say:
http://gcc.gnu.org/ml/gcc-testresults/2006-11/msg00519.html
and try and replicate the result. If you can, then you probably
don't have to worry about your system. If you can't... well, not
much we can probably do.
If you can't find any results to compare to, you might want to use
cygwin and compare against, say
http://gcc.gnu.org/ml/gcc-testresults/2007-02/msg00886.html
or http://gcc.gnu.org/ml/gcc-testresults/2007-01/msg00811.html
*sigh*
http://gcc.gnu.org/ml/java-patches/2007-q1/msg00321.html
The apparent problem, as I know I made clear, is not specific to my
system or MGW. The underlying problem, if there is one, is probably in
the optimizing code, but I don't know enough to say that with any
certainty.
The point is, if it is two different problems, one MGW specific
problem and one other problem, manifesting very similar issues than
reporting it to GCC instead of MGW would be a... problem.
Soma
