#5 Fails to configure on ia64 architecture

[Sam Couter]

I recieved the following bug report from one of the
people compiling Debian packages for other
architectures. I don't understand how the configure
script is determining the alignment requirements, so I
can't fix it myself.

As a Debian developer, I have access to an ia64
machine, so I can try to run and test code developed by
someone else. Inconvenient, yes, but I don't possess
the necessary skills and knowledge to do it myself.
Maybe if someone explains it to me I can do it.

Anyway, Debian bugreport follows:
---

Package: njamd
Version: 0.9.2-3
Severity: important

Building njamd on ia64 fails when the configure script
is for the default alignment requirement. Here are the
last few lines up to the failure:

| checking for snprintf... yes
| checking for select... yes
| checking for strdup... yes
| checking for strerror... yes
| checking for strtol... yes
| checking for backtrace_symbols_fd... yes
| checking for mmap... yes
| checking processor bitwidth... 64
| checking default alignment requirement... configure:
error: Mike Perry is a moron. Mail mikepery@fscked.org
and tell him his alignment check does not work.
| make: *** [configure-stamp] Error 1

The full build log is availible here if you are interested:
http://people.debian.org/~dsp/buildlog/njamd_0.9.2-3_1003857822

[Sam Couter]

Logged In: YES
user_id=14486

This patch was submitted for the Debian package by John R
Daily <jdaily@progeny.com>.

[Sam Couter]

Patch for 64 bit systems

[Stäphen Goedzal]

Both before and after applying the above patch:

configure:3375: checking If we can trust gcc with bitfields
configure:3409: gcc -o conftest -g -O2 -Wall -I/usr/local/src/njamd-0.9.3pre2/njamd-0.9.3pre2/include conftest.c -lnsl 1>&5
configure:3386:5: error: width of 'y' exceeds its type
int y : sizeof(long)*8 - 4;
^

All hunks succeded:
patching file configure.in
Hunk #1 succeeded at 201 (offset 65 lines).
Hunk #2 succeeded at 213 (offset 65 lines).
Hunk #3 succeeded at 225 (offset 65 lines).
Hunk #4 succeeded at 237 (offset 65 lines).
Hunk #5 succeeded at 268 (offset 65 lines).
Hunk #6 succeeded at 280 (offset 65 lines).
Hunk #7 succeeded at 305 (offset 65 lines).
Hunk #8 succeeded at 319 (offset 65 lines).
Hunk #9 succeeded at 333 (offset 65 lines).
patching file include/fe/util/libcwrap.h
patching file src/fe/console/console.c
patching file src/fe/util/libcwrap.c
patching file tests/alignment.c
patching file tests/dltest.c
patching file tests/postfree.c
patching file tests/signal.c
Hunk #1 succeeded at 1 with fuzz 1.

Note that the error configure produced when run was slightly different from above. I ran it directly from the 0.9.3pre2 tarball in the files section of this project..

See attached config.log for details (search for "trust gcc").