Mutating requires file

[Adrian Buehlmann]

On 01.07.2008 13:22, Adrian Buehlmann wrote:
> Is on the fly adding of entries to the .hg/requires file a good thing?
> Is it wanted?
> Or just tolerated in some cases?
> 
> In his patch, Jesse only adds "longnames" to .hg/requires *if* a long
> filename is actually encountered (e.g. during pull, commit, ..).
> 
> Should that concept be extended to auxencode [1], that is,
> adding "auxencode" on the fly to .hg/requires as soon as a filename
> that needs to be auxencoded is encountered?
> 
> Or should we just combine "auxencode" and "longnames" into one
> static requires entry which implies both ("winquirks")?
> 
> Should "winquirks" be added to .hg/requires as as soon as a new repo
> is created, even on unix?
> 
> [1] http://www.cadifra.com/cgi-bin/repos/hg-auxencode/file/tip/auxencode.patch
> 

After having some chat on IRC (djc, tonfa, ronny, parren),
I've come to the conclusion that mutating requires files entries
are really not worth the trouble.

Simply writing "winquirks" (or whatever that name will be) to the requires
file for newly created (or newly cloned) repos is good enough.

The vast majority of the repos spread over the wire anyway, so
encountering a new repo with an old Mercurial should be rather rare,
especially on unix. Such a repo would have been transferred on the
file system level, which mostly happens in rather controlled circumstances
anyway (mostly backup, rather local).

Also, bundles are unaffected just like the wire format.

Let's drop "mutating requires files".