X-Authentication-Warning: delorie.com: mail set sender to djgpp-bounces using -f X-Authentication-Warning: itservs.wilkes.edu: apache set sender to fdonahoe AT wilkes DOT edu using -f Message-ID: <1095337866.4149878a67850@webmail.wilkes.edu> Date: Thu, 16 Sep 2004 08:31:06 -0400 Disposition-Notification-To: fdonahoe AT wilkes DOT edu X-Confirm-Reading-To: fdonahoe AT wilkes DOT edu X-pmrqc: 1 From: fdonahoe AT wilkes DOT edu To: Andris Pavenis Cc: djgpp AT delorie DOT com Subject: Re: bnu214s compiled for djgpp 2.04 alpha MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit User-Agent: Internet Messaging Program (IMP) 3.2 X-Originating-IP: 146.94.1.210 I should say two things first. 1. The binaries of later versions than appear on Andrew Cottrell's page, and all of the sources have been downloaded from current/v2gnu on delorie.com. 2. Maybe I'm not working under DOS after all. The window is opened by a shortcut to "comd.exe" for the simple reason that I could not get doskey to work in a "command.com" window. Quoting Andris Pavenis : >> .... > Some notes: > > bash205b[bs].zip from clio.rice.edu has one critical bug, > which prevents running rather many configure scripts (when > PATH_SEPARATOR=:). See my recent post to djgpp-workers > mailing list for more details: > http://www.delorie.com/djgpp/mail-archives/browse.cgi?\ > p=djgpp-workers/2004/09/06/09:13:35 > > Also there is one problem in config.site which appears when > environment variable PATH_SEPARATOR is not set at all. > Thanks! I have not yet downloaded bash sources. I have a slow telephone connection. > ... > I don't remember recent version of autoconf and automake > which passes all tests for DJGPP target. autoconf-2.59 and > automake-1.9.1 has even more failures, but seems to work > for me otherwise. > autoconfig-2.57 passed all the tests after its build. > ... > I have practically not used UPX for some time. > > upx-1.2.X had problems with DJGPP binaries generated by > recent versions of GCC and binutils. Really the failure > is if one decompress UPX compressed file (it does not work > any more). Therefore last versions gives up on DJGPP > binaries. New beta versions of UPX (eg. upx-1.9.2 works, > but it's not tested so much, so be carefull) > The problems with the binutl-2.14 script were really trivial. Warnings about the age of autoconfig and a failure to produce a zipped up binary package for distribution. UPX worked because it found a djgpp2/coff format. Unexpected because it hadn't worked on the executables from the other packages. Frank Donahoe wrote > > I'd like to rebuild configure, Andris, if you could walk me > > through the steps. > It may depend from package. In some cases aclocal, automake > mey need some command line options. Usually they should work > without > > > Do I go down to maintainer-clean? > Normally not necessary > > Andris > There were other more disturbing incidents. The package, dif281s built, with no problems, out of the box but did not use NLS support. It had to be reconfigured to provide this option. configure was built with autoconfig=2.53. It produced two files which were symlink include files (.h). The type of error is similar to that illustrated here from the gmp-4.12 package. gcc -DHAVE_CONFIG_H -I. -I. -I.. -D__GMP_WITHIN_GMP -I.. -DOPERATION_mp_bases -O2 -fomit-frame-pointer -mcpu=i486 -march=i386 -c mp_bases.c `-mcpu=' is deprecated. Use `-mtune=' or '-march=' instead. In file included from ../gmp-impl.h:39, from mp_bases.c:31: ./gmp-mparam.h:1: error: parse error before '!' token In file included from ../gmp-impl.h:39, from mp_bases.c:31: ./gmp-mparam.h:2:478: warning: no newline at end of file In file included from mp_bases.c:31: ./gmp-impl.h:190: error: parse error before '}' token Parse error before '!' token! It was the first character in the file. "ls -l" revealed that it was indeed a file of 510 bytes, but it was a dead link! If I hadn't read Richard Dawe's discussion of the possible errors testing automake-1.76 and what he had done to get one test to work, I should never have thought to apply "dtou" to the dead links. They became live links and "ls" now reported the size of the file which was linked rather than that of the symlink file. Unfortunately I don't remember and didn't make a record of the failure with dif281s. So I can't report here the names of the two symlink files. I do have a record of the successful trial which followed the revival of the symlinks. I will return to the gmp412s package later. There is much more trouble there than a score of dead symlinks. Frank ----------------------------------------------------------------------- This mail was sent through Wilkes Webmail: http://webmail.wilkes.edu Wilkes Webmail is using IMP: http://horde.org/imp/