On Fri, Jan 25, 2002 at 10:06:30AM -0800, Michael A Chase wrote: >----- Original Message ----- >From: "Christopher Faylor" <[EMAIL PROTECTED]> >To: <[EMAIL PROTECTED]> >Sent: Friday, January 25, 2002 09:24 >Subject: Re: [PATCH]Package extention recognition (revision 2) > > >>On Fri, Jan 25, 2002 at 02:59:17AM -0800, Michael A Chase wrote: >>>And that test is still there, I moved it into the if () so something >>>like ".tar.bz2" wouldn't trigger the return .... : 0; If all the ifs >>>fail, return 0; still occurs. >> >>Hmm. Seems like someone has "improved" this code from when I wrote it. > >Is the "improved" version my change or the way I found it?
The current version, prior to your change has been "improved". >>My version checked for a trailing component. If it existed, it >>returned the index into the string. >> >>This version sort of does the same thing but if there is a .tar.bz2 >>anywhere in the string prior to trailing component, it will fail >>regardless of whether the filename ends with .tar .tar.gz or .tar.bz2. >> >>Perhaps that is an acceptable risk but it puzzles me why anyone would >>move from an algorithm that was foolproof to one that wasn't. > >I can go either way. It is hard for me to imagine >foo-0.0.tar.bz2.tar.gz being valid, but my patched version would accept >it while the original version would reject it. Ok. I didn't look closely at your patch. My version would accept the above, too. >Neither version is fool proof for some values of fool; both would pass >foo-0.0.tar.gz.tar.bz2. If you like I can move the '(end - ext) == x' >test back to the 'return' statement. I guess that's up to Robert. The ChangeLog indicates that he made this change. cgf