Обсуждение: [GENERAL] testcase failing on git master / how to progress
Hi, I just started last week to build postgresql from source. (So this may well be something I did wrong on my side) I am on branch master Revision: f7819baa618c528f60e266874051563ecfe08207 Date: 17/03/2017 18:58:06 I did build with make clean distclean ./configure --prefix=/deploys/postgresql/inst/ --with-pgport=5433 --enable-debug CFLAGS="-ggdb -O0 -g3 -fno-omit-frame-pointer" make make install And I got a few tests failing. macaddr8 ... FAILED opr_sanity ... FAILED object_address ... FAILED If I build with ./configure --prefix=/deploys/postgresql/inst/ --with-pgport=5433 then all tests pass. So questions: - Can anyone else reproduce this? - How to establish if this is a problem with my build, or a bug in pg? (And if the latter, what next) ____ one more note: make check has a problem on my system, but I (believe I) have worked around. I can see that "make check" sets PATH="/deploys/postgresql/postgresql/tmp_install/deploys/postgresql/inst/bin:$PATH" LD_LIBRARY_PATH="/deploys/postgresql/postgresql/tmp_install/deploys/postgresql/inst/lib" but this path does not exist, and therefore binaries are not found. I did install the build, and set the path (and ld-path) to include the install location, then run the tests. I have not found what causes this issue, but maybe it is the --prefix? uname -a FreeBSD bsd1 10.0-RELEASE FreeBSD 10.0-RELEASE #0 r260789: Fri Jan 17 01:46:25 UTC 2014 root@snap.freebsd.org:/usr/obj/usr/src/sys/GENERIC i386 gmake --version GNU Make 4.1 Built for i386-portbld-freebsd10.0 cc --version FreeBSD clang version 3.3 (tags/RELEASE_33/final 183502) 20130610 Target: i386-unknown-freebsd10.0 Thread model: posix
On 03/17/2017 06:07 PM, Martin F wrote: > Hi, > I just started last week to build postgresql from source. (So this may > well be something I did wrong on my side) Do you really want to build the latest dev version? If not and you want a specific version you should look here: https://www.postgresql.org/ftp/source/ > > I am on branch master > Revision: f7819baa618c528f60e266874051563ecfe08207 > Date: 17/03/2017 18:58:06 > > I did build with > > make clean distclean > ./configure --prefix=/deploys/postgresql/inst/ --with-pgport=5433 > --enable-debug CFLAGS="-ggdb -O0 -g3 -fno-omit-frame-pointer" > make > make install > > And I got a few tests failing. > macaddr8 ... FAILED > opr_sanity ... FAILED > object_address ... FAILED > > If I build with > ./configure --prefix=/deploys/postgresql/inst/ --with-pgport=5433 > then all tests pass. > > So questions: > - Can anyone else reproduce this? > - How to establish if this is a problem with my build, or a bug in pg? > (And if the latter, what next) > > ____ > one more note: > make check > has a problem on my system, but I (believe I) have worked around. > I can see that "make check" sets > PATH="/deploys/postgresql/postgresql/tmp_install/deploys/postgresql/inst/bin:$PATH" > LD_LIBRARY_PATH="/deploys/postgresql/postgresql/tmp_install/deploys/postgresql/inst/lib" > > but this path does not exist, and therefore binaries are not found. > I did install the build, and set the path (and ld-path) to include the > install location, then run the tests. > I have not found what causes this issue, but maybe it is the --prefix? > > uname -a > FreeBSD bsd1 10.0-RELEASE FreeBSD 10.0-RELEASE #0 r260789: Fri Jan 17 > 01:46:25 UTC 2014 root@snap.freebsd.org:/usr/obj/usr/src/sys/GENERIC i386 > > gmake --version > GNU Make 4.1 > Built for i386-portbld-freebsd10.0 > > cc --version > FreeBSD clang version 3.3 (tags/RELEASE_33/final 183502) 20130610 > Target: i386-unknown-freebsd10.0 > Thread model: posix > > > > -- Adrian Klaver adrian.klaver@aklaver.com
On 18/03/2017 01:22, Adrian Klaver wrote: > On 03/17/2017 06:07 PM, Martin F wrote: >> Hi, >> I just started last week to build postgresql from source. (So this may >> well be something I did wrong on my side) > > Do you really want to build the latest dev version? > Yes, I want the latest, and I understand, that it may be unstable. this is not using the postgres in any production/development. Rather that I want to start exploring the postgres sources themself.
Martin F <pg@mfriebe.de> writes: > I did build with > make clean distclean > ./configure --prefix=/deploys/postgresql/inst/ --with-pgport=5433 > --enable-debug CFLAGS="-ggdb -O0 -g3 -fno-omit-frame-pointer" > make > make install That looks like it ought to work. Did you get any compile warnings? (Personally I tend to do the make steps as "make -s" so that you can spot the warnings if there are any.) > And I got a few tests failing. > macaddr8 ... FAILED > opr_sanity ... FAILED > object_address ... FAILED Hard to say much about that. Looking at the actual regression.diffs might be more informative. A general tip when playing around with development HEAD is to look at the buildfarm: https://buildfarm.postgresql.org/cgi-bin/show_status.pl If critters are falling over left and right on the commit you pulled, then you got a bad version. If all is green on the buildfarm but it's not working for you, then we'd like to hear details. regards, tom lane
On Sat, Mar 18, 2017 at 2:06 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote: > If critters are falling over left and right on the commit you pulled, > then you got a bad version. If all is green on the buildfarm but > it's not working for you, then we'd like to hear details. If you can still produce failures even on latest HEAD, and that the environment you are testing is more particular than anything currently on the buildfarm, creating a new buildfarm animal would be a good idea as your settings would be immediately covered and failure would be tracked easily. -- Michael