On 11/10/2011 04:40 PM, C. Michael Pilato wrote:
On 11/10/2011 10:29 AM, Neels J Hofmeyr wrote:
It seems to me that excluding only those externals (dir& file) that are
fixed to a specific revision is the best solution. My only worry are all
those users out there expecting dir externals to be excluded always.
That's why I'm asking: if I told everyone to place a specific revision in
their externals definitions to be able to exclude them from commits, would
that cause major havoc?
Major havoc? Perhaps not so much. But realize that we'd be telling folks
to make a versioned change to *their data* solely for the purpose of
preserving a behavior they have grown to expect already. We're not asking
them to tweak local configuration, or some process point -- we're asking
them to change their repository contents. That starts to feel (to me, at
least) like we've crossed a line we shouldn't cross.
Where does this argument come in with file externals? Current trunk changes
the default behavior compared to 1.7.x, so that file externals are no longer
committed along by default. Would you want to revert that change?
"Fixing" externals is a dilemma at this juncture. In the end, we should have
a sane default with all the ways available to configure otherwise;
unfortunately this *will* force users to adjust, either way.
We can also: never include externals by default (current trunk), with new
svn:externals syntax to flag certain externals to be included in commit
recursion (stsp's suggestion). That, too, would allow 1.7.x behavior.
Maybe a little survey on users@ could help ... or maybe that would confuse
us for good ;)
~Neels
