Обсуждение: three VERY minor things with 7.1 final
I want to thank you for the excellent and fast responses I have received in the past. Especially while troubleshooting the sfio problem. These problems are VERY minor and easily worked around. Part of the reason I am posting them is just in case someone else runs across the same things. I am running Sparc Solaris 7, GCC 2.95.3. Configure line was: ./configure --prefix=/usr/local --with-perl --enable-odbc --enable-syslog. I just built the postgresql 7.1 final and the configure script is still checking for sfio. Not a major big deal, but I need to remove the sfio check from configure.in, run autoconf, and then configure to fix it. If you remember sfio on Sparc Solaris 7 was causing a segfault when psql would display its output. Also, when I create a table and actually name the primary key using the following: create table fcos ( "FeatureNumber" integer NOT NULL, "FeatureName" varchar(32), "Feature1" integer DEFAULT 0, CONSTRAINT "PK_fcos" PRIMARY KEY ("FeatureNumber") ); I get the following: NOTICE: CREATE TABLE/PRIMARY KEY will create implicit index 'PK_fcos' for table 'fcos' This is really a nit-pick but I am explicitly naming it so why should I get a notice about it? Also, during make install I am getting "lack of permissions" when it tries to install Perl. I have Perl 5.6.0 on this machine built from source. It tells me to fix the problem and go into src/interfaces/perl5 and do a make install. Without fixing any problem I just go into src/interfaces/perl5 and do a make install and it works without returning any errors. BTW, I am doing the make install as root and all of the directory (both source and destination) have reasonable priviledges.
David George <david@onyxsoft.com> writes: > I just built the postgresql 7.1 final and the configure script is still > checking for sfio. Not a major big deal, but I need to remove the sfio > check from configure.in, run autoconf, and then configure to fix it. If > you remember sfio on Sparc Solaris 7 was causing a segfault when psql > would display its output. Hmm. Digging in the CVS logs and pghackers archives, it seems that configure started checking for sfio in response to Michael Richards' unsubstantiated claim that FreeBSD 2.2.5 requires it to be used (http://www.postgresql.org/mhonarc/pgsql-hackers/1998-04/msg00363.html). Can anyone confirm or deny that? Marc had previously expressed some interest as well (thread starting at http://www.postgresql.org/mhonarc/pgsql-hackers/1998-04/msg00219.html) but the outcome of that thread didn't seem to suggest that it's a must-have item. I'm tempted to rip out the configure check for sfio, but if it's only broken on some systems and really is useful on others, then we have to try to figure out how to tell if it's broken :-( > Also, during make install I am getting "lack of permissions" when it > tries to install Perl. I have Perl 5.6.0 on this machine built from > source. It tells me to fix the problem and go into src/interfaces/perl5 > and do a make install. Without fixing any problem I just go into > src/interfaces/perl5 and do a make install and it works without > returning any errors. No clue about this. The makefile is testing for writability of perl's INSTALLSITELIB, which ought to succeed if you are root. Besides which, if you go into that directory and do "make install" again, it ought to fail again in the same way. Unless maybe root's path is different, and you are invoking make not gmake the second time? regards, tom lane
On Tue, 17 Apr 2001, Tom Lane wrote: > David George <david@onyxsoft.com> writes: > > I just built the postgresql 7.1 final and the configure script is still > > checking for sfio. Not a major big deal, but I need to remove the sfio > > check from configure.in, run autoconf, and then configure to fix it. If > > you remember sfio on Sparc Solaris 7 was causing a segfault when psql > > would display its output. > > Hmm. Digging in the CVS logs and pghackers archives, it seems that > configure started checking for sfio in response to Michael Richards' > unsubstantiated claim that FreeBSD 2.2.5 requires it to be used > (http://www.postgresql.org/mhonarc/pgsql-hackers/1998-04/msg00363.html). > Can anyone confirm or deny that? Marc had previously expressed some > interest as well (thread starting at > http://www.postgresql.org/mhonarc/pgsql-hackers/1998-04/msg00219.html) > but the outcome of that thread didn't seem to suggest that it's a > must-have item. Damn, now *that* is an old thread ... I can't see any reason why it would be required for FreeBSD, as its only a port for the OS, not part of it ... > I'm tempted to rip out the configure check for sfio, but if it's only > broken on some systems and really is useful on others, then we have to > try to figure out how to tell if it's broken :-( Pull it, as far as I'm conerned ...
Tom Lane writes: > I'm tempted to rip out the configure check for sfio, but if it's only > broken on some systems and really is useful on others, then we have to > try to figure out how to tell if it's broken :-( I just installed sfio here and built PostgreSQL with it and didn't see any difference. Which is not surprising because libsfio doesn't define any symbols that PostgreSQL uses. (It's a completely separate interface, all the functions are named sf*.) What you'd really need to use is libstdio (plus libsfio), which is a wrapper. So I removed the check in configure because it obviously had a net negative benefit. Those who want to use it can still add '-lstdio -lsfio' to LIBS. -- Peter Eisentraut peter_e@gmx.net http://funkturm.homeip.net/~peter