timestamps, take N+1

[Ben Collins-Sussman <su...@collab.net>]

OK, I'm bringing up the timestamps feature one more time.  Everyone
brace yourselves.  :-)

A quick recap:

  * Absolutely everyone wants 'svn export' to use commt-timestamps.
    This will be done, it's just an itty-bitty change to the export
    editor.  Very easy, end of story.

  * Some folks want svn to have a complete mtime preservation system.
    Some folks don't like this idea.  But either way, it involves
    creating new property types, new marshalling, and from what I can
    tell, is generally way too complex to just "throw into" svn 1.0.

  * Some folks want svn to have cvs-like behavior:  that is, have 'svn
    checkout' to set commit-timestamps.  Some variations have been
    discussed that have 'svn up' either [always|sometimes] using
    current timestamps.  And some people don't like any of this idea,
    they find it to be inconsistent/weird, and think the status quo
    (to do nothing) is just perfect.

I have a concrete proposal now, here it is:

  1. Have 'svn export' use commit-timestamps unconditionally.

  2. Add a switch to both 'svn checkout/update' that causes all
     affected files to receive commit-timestamps.  By default, these
     commands continue to use 'now' times as they do already.


Here's why I like this proposal:

  * gives users a choice.  By default, most developers want 'now'
    timestamps, so that 'make' works as they expect.  But release
    managers or automated build systems can pass the new switch, and
    get the commit-times they want.

  * tiny code change.  We already have the commit-timestamps available
    in update_editor.c; it's just a matter of adding a flag to the
    edit_baton and tweaking svn_wc_install_file().

  * matches a cvs feature.  Our target audience is CVS users, and a
    lot of them have come to expect/depend on the availability of
    commit timestamps.  Some may argue that svn is more "sane" by not
    making it the default, but at least the feature will be available.



---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: timestamps, take N+1

[Garrett Rooney <ro...@electricjellyfish.net>]

Ben Collins-Sussman wrote:

> I have a concrete proposal now, here it is:
> 
>   1. Have 'svn export' use commit-timestamps unconditionally.
> 
>   2. Add a switch to both 'svn checkout/update' that causes all
>      affected files to receive commit-timestamps.  By default, these
>      commands continue to use 'now' times as they do already.

+1

This seems like the best way to go.

-garrett


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: timestamps, take N+1

["P.Marek" <pm...@users.sourceforge.net>]

> OK, I'm bringing up the timestamps feature one more time.  Everyone
> brace yourselves.  :-)
>
[snip]
>
> I have a concrete proposal now, here it is:
>
>   1. Have 'svn export' use commit-timestamps unconditionally.
>
>   2. Add a switch to both 'svn checkout/update' that causes all
>      affected files to receive commit-timestamps.  By default, these
>      commands continue to use 'now' times as they do already.
That ok for now.

But it won't meet my criteria. I need to have the *original* timestamps 
versioned, as they are the most acknowledged attribute I have to 
differenciate between versions (the external source's versions, not my 
svn-internal)

It's fine with me if that's used now in svn - but if you want to have the 
files' mtime instead of commit-times there'll have to be a new property (and 
maybe property class), which has to be special cased in some functions, ...
So I believe that most of the code needed comes from this one change - use 
mtime instead of commit-time.

Time will show ...



---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: timestamps, take N+1

[Ben Collins-Sussman <su...@collab.net>]

Branko Čibej <br...@xbc.nu> writes:

> Ben Collins-Sussman wrote:
> 
> >OK, I'm bringing up the timestamps feature one more time.  Everyone
> >brace yourselves.  :-)
> >
> brrrr...
> (I hope you don't mind if I just use half a brace :-)
> 
> >I have a concrete proposal now, here it is:
> >
> >  1. Have 'svn export' use commit-timestamps unconditionally.
> >
> >  2. Add a switch to both 'svn checkout/update' that causes all
> >     affected files to receive commit-timestamps.  By default, these
> >     commands continue to use 'now' times as they do already.
> >
> 
> Why not be consistent and always use an option, even at export? -0 on
> "svn export" changing behaviour unconditionally.

Fine with me.  I'll have co/up/export all take the new option.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org