Testing xgpc 20030830, based on gcc-3.3.2 (i686-pc-linux-gnu) (srv1.vchs.cz), 2003-10-18 18:54:06 echo "/home/dave/gcc-obj/gcc/xgpc -B/home/dave/gcc-obj/gcc/ -g -O3 -W -Wall -Wundef -Wpointer-arith -Wno-unused " /home/dave/gcc-obj/gcc/xgpc -B/home/dave/gcc-obj/gcc/ -g -O3 -W -Wall -Wundef -Wpointer-arith -Wno-unused PC="/home/dave/gcc-obj/gcc/xgpc -B/home/dave/gcc-obj/gcc/" PFLAGS="--autobuild -g -O3 -W -Wall -Wundef -Wpointer-arith -Wno-unused -I ../rts --unit-path=/home/dave/gcc-obj/gcc/p/units " SRCDIR="/home/dave/gcc-3.3.2/gcc/p/test" TEST_MAKE_FLAG=test-make-flag "/home/dave/gcc-3.3.2/gcc/p/test/test_run" | tee test_log | "/home/dave/gcc-3.3.2/gcc/p/test/test_sum" | tee test_summary TEST agettext2test.pas: SKIPPED: could not set locale `de_DE' TEST fjf105a.pas: SKIPPED: librx not installed TEST fjf105b.pas: SKIPPED: librx not installed TEST fjf165a.pas: SKIPPED: German locale not installed TEST regextest.pas: SKIPPED: librx not installed

# of GPC tests 3794 # of GPC tests passed 3789 # of GPC tests skipped 5 # of GPC tests failed 0 make[3]: Leaving directory `/home/dave/gcc-obj/gcc/p/test'

It is great, isn`t it ?

Regards, David

On Sat, 18 Oct 2003, David Kredba wrote:

When do U downloaded your copy of gpc release file ?

It contains a few fixes, I think, added after the first release.

It wasn't the original 2003080 release, but I see the length is not the same as todays. So downloaded today's version and started over.

(begin rant) Swell. If one can't rely on the date code in the title what's the use of having it? Might as well just call it "gpc.tar.gz" and be done with it. (end rant)

Configure params : ./gcc-3.2.3/configure --prefix=/usr/local --enable-shared --enable-threads=posix --disable-checking --enable-long-long --enable-__cxa_atexit --enable-languages=c,c++,ada,f77,objc,java,pascal --with-system-zlib

configure here was shorter: ../gcc-3.3.2/configure --enable-languages=pascal --enable-threads=posix --enable-__cxa_atexit

results: tests 3794 passed 3791 skip 3 fail 0

Thanks, Russ

Frank Heckenbach wrote:

Conclusion (as far as I'm concerned):

• Never make a release under pressure anymore. (Everyone, if you want to tell me why a new release is urgently needed at any time, please redirect to /dev/null.)

• Never do post-release updates, even if serious bugs. (Except perhaps for the proposed alpha/beta scheme.)

Frank, I was surprised that the updates have the same version number as original release, and that is the repeating complaint. I do not know what your release scripts do, but AFIKS just renaming tarballs and the main source directory would completely avoided any confusion. If done by hand, such renaming should take no more than one hour work (that is conservative estimate, assuming that you would have to manually transfer files and type in passwords and then fetch back the release to check it). I estimate that such renaming would take less time then answering resulting e-mails.

The point is: never change a release, it is much better to fake a new release (if you want to avoid your normal release procedure).

Frank, please do not take that remarks too personally. You are doing very good job maintaining GPC. This is just a little detail in contrast with "best practice". And had you wrote "I will rename the relase" I would not write this message.

echo "I would be nice to have a new relase (even if it contains" \ "nothing new), just to make clear that all stale copies of" \ "gpc-20030830.tar.gz are obsolete" > /dev/null

Russell Whitaker wrote:

Suggestion: when you upload a new tarball that has the same version number change the name of the tarball to add a revision number. thus the third revision of gpc-20030830 would be gpc-20030830-r3.tar.gz and the untared tree would still be gpc-20030830/

Then I'll probably get the next flames from those who use auto-builder scripts that expect certain file names ...

Frank