Date: 2008-08-05 13:39:25
On Tue, 5 Aug 2008, John Maddock wrote:
> Sure, but somehow you've ended up with mixed \n and \r\n line endings, and
> that's really not something you should be checking in IMO. As for why the
> file ended up that way, that's a whole other issue...
Official or not, Boost effectively endorses \n as the one true line
ending; the release files have names like boost_1_35_0.7z and
boost_1_35_0.tar.bz2, no line-ending choice available.
SVN should not act like FTP with text and binary modes, munging bytes as
they flow. It should maintain bytewise-exact copies of what the user
intended, possibly with commit hooks to reject bad line endings.
P.S. Once you have exact copies, you can add a layer to cryptohash each
revision, checking for corruption in the system... Each patch would
contain the hash of the tree it applies to. Then you could use those
cryptohashes for distributed revision control... ;)
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk