Steve Langasek wrote: > Package: libdbi-drivers > Version: 0.8.2-1-4 > Severity: serious > Tags: sid patch > > The libdbi-drivers package fails to build with freetds 0.82 in unstable, > because the upstream checks for libraries (in acinclude.m4) are absurd: > > if test "$ac_freetds_incdir" = "no" || test "$ac_freetds_libs" = > "no"; then > freetds_incdirs="/usr/include /usr/local/include" > AC_FIND_FILE(tds.h, $freetds_incdirs, ac_freetds_incdir) > freetds_libdirs="/usr/lib /usr/local/lib" > AC_FIND_FILE(libtds.so, $freetds_libdirs, ac_freetds_libdir) > if test "$ac_freetds_incdir" = "no"; then > AC_MSG_RESULT(no) > AC_MSG_ERROR([Invalid Freetds directory - include > files not found.]) > fi > if test "$ac_freetds_libdir" = "no"; then > AC_MSG_RESULT(no) > AC_MSG_ERROR([Invalid Freetds directory - libraries > not found.]) > fi > fi > > AC_FIND_FILE() is the wrong thing. We have compilers with system lib and > include paths, and autoconf macros AC_CHECK_HEADERS() and AC_CHECK_LIB() > which do the right thing with the compiler. AC_FIND_FILE() here is > reinventing the wheel - poorly. > > Besides which, this check uses AC_FIND_FILE() to look for a file *which is > never used in the build*. libtds.so /was/ an internal library in previous > versions of FreeTDS, which has now been removed. libdbi-drivers /never/ > linked against this library; it linked against libct, which is the correct > public interface. So it should be checking for the library it will actually > use! (but using AC_CHECK_LIB(), not AC_FIND_FILE()!) > > There is fortunately a way to bypass this wrong check by passing a > --with-freetds-dir argument to ./configure. A patch to do this is attached. > Let me know if you would like me to upload this as an NMU. > > Thanks, > > > ------------------------------------------------------------------------ > > diff -u libdbi-drivers-0.8.2-1/debian/rules > libdbi-drivers-0.8.2-1/debian/rules > --- libdbi-drivers-0.8.2-1/debian/rules > +++ libdbi-drivers-0.8.2-1/debian/rules > @@ -20,7 +20,10 @@ > cp /usr/share/misc/config.sub . > cp /usr/share/misc/config.guess . > > - ./configure --host=$(DEB_HOST_GNU_TYPE) --build=$(DEB_BUILD_GNU_TYPE) > --prefix=/usr --with-pgsql --with-mysql --with-sqlite --with-sqlite3 > --with-freetds --enable-docs --enable-libdbi > + ./configure --host=$(DEB_HOST_GNU_TYPE) --build=$(DEB_BUILD_GNU_TYPE) \ > + --prefix=/usr --with-pgsql --with-mysql --with-sqlite \ > + --with-sqlite3 --with-freetds --with-freetds-dir=/usr \ > + --enable-docs --enable-libdbi > # This is for the moment removed until the upstream fixes the problems with > # the new API: --with-firebird > > diff -u libdbi-drivers-0.8.2-1/debian/changelog > libdbi-drivers-0.8.2-1/debian/changelog > --- libdbi-drivers-0.8.2-1/debian/changelog > +++ libdbi-drivers-0.8.2-1/debian/changelog > @@ -1,3 +1,13 @@ > +libdbi-drivers (0.8.2-1-4.1) unstable; urgency=high > + > + * Non-maintainer upload. > + * High urgency upload for RC bugfix. > + * Pass --freetds-inc-dir to configure, to bypass the ridiculous upstream > + AC_FIND_FILE() checks for files that *aren't used during the build*; > + fixes a build failure with FreeTDS 0.82. > + > + -- Steve Langasek <[EMAIL PROTECTED]> Sat, 02 Aug 2008 09:00:11 +0000 > + > libdbi-drivers (0.8.2-1-4) unstable; urgency=low > > * Updated the dbd_pgsql.c so it doesn't conflicts with the declaration of
Hi, I'm not against an NMU, I'm pretty loaded with work. Does that bug appears as well on Lenny? If so, can you manage to contact debian-release to make them send the patched version to Testing? Thomas -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
