Insecure temporary file usage in developer/build tools

Hi PostgreSQL developers!

Debian's security audit team recently reviewed PostgreSQL for insecure
temporary file usage and found that a lot of the developer tools and
also some build tools are vulnerable against symlink attacks.

Please see

http://bugs.debian.org/291962

for the detailled report and a proposed patch. However, the patch
mostly relies on mktemp and/or tempfile, which might not be available
on all platforms that PostgreSQL supports, so the patch should be
regarded as a pointer to the issues, not as their solution.

In many cases the usage of a temporary file is not necessary in the
first place and can be replaced by piping (as I did with
src/test/bench/perquery) or "OUTPUT=$(program args)" constructs. If
neither is possible, then the script should at least be run with "set
-o noclobber" (or "#!/bin/sh -C", which does the same).

Apart from the Debian-specific issues of the bug report (which I
already fixed), the following files were found to be vulnerable:

Used in build, fixed for Debian (see attached patch):
postgresql-7.4.6/src/backend/catalog/genbki.sh
postgresql-7.4.6/src/test/bench/perquery

Not shipped in the Debian package, since it is useless:
postgresql-7.4.6/contrib/pg_upgrade/pg_upgrade

Not used anywhere in the source, not fixed in Debian package:
postgresql-7.4.6/src/include/catalog/duplicate_oids
postgresql-7.4.6/src/tools/ccsym
postgresql-7.4.6/src/tools/find_static
postgresql-7.4.6/src/tools/make_ctags
postgresql-7.4.6/src/tools/make_etags
postgresql-7.4.6/src/tools/pgtest
postgresql-7.4.6/src/tools/pginclude/pgcompinclude
postgresql-7.4.6/src/tools/pginclude/pgdefine
postgresql-7.4.6/src/tools/pginclude/pgfixinclude
postgresql-7.4.6/src/tools/pginclude/pgrminclude
postgresql-7.4.6/src/tools/pgindent/pgcppindent
postgresql-7.4.6/src/tools/pgindent/pgindent
postgresql-7.4.6/src/tools/pgindent/pgjindent
postgresql-7.4.6/contrib/tools/add-emacs-variables

Most of the tools are irrelevant for the sake of package building and
shipping, but they might be commonly used on PostgreSQL developer's
machines, so it might be a good idea to fix then eventually.

Thanks for considering and for your great work and have a nice day!

Martin

--
Martin Pitt http://www.piware.de
Ubuntu Developer http://www.ubuntulinux.org
Debian GNU/Linux Developer http://www.debian.org