Thomas Bushnell BSG <tb@becket.net> writes: > On Wed, 2006-12-27 at 09:36 +0000, Marc 'HE' Brockschmidt wrote: >> Thomas Bushnell BSG <tb@becket.net> writes: >>> And one seems perhaps to be responsible for a regression in gnucash >>> (see #404585). >> Yes, this is due to a stricter input validation in gkeyfile.c. The >> validation is an ugly C replacement for this: >> m!^[-_/+.[:alnum:]]+(\[[-_.@[:alnum:]]+\])?$! > > I'm uncertain how to interpret this message. You said that the new > release only fixes bugs, and that you reviewed all the changes, and > that the new release includes a change in implementation from a regexp > to a C replacement, which more strictly validates input. What? No, what I meant is that input validation was introduced to fix other problems. [1] As I'm not keen on pasting 20 lines of boring C routines, I created an equivalent (perl) regex to demonstrate what Input is currently allowed. As this may break more applications (earlier version broke locale parsing and gnomevfs), we should probably keep that code, reduce it to a warning for etch and then work out (together with upstream) how to solve this for the future. Marc Footnotes: [1] http://bugzilla.gnome.org/show_bug.cgi?id=343191 -- BOFH #114: electro-magnetic pulses from French above ground nuke testing.
Attachment:
pgpLodPZZj9Eh.pgp
Description: PGP signature