Re: Import performance with David Kimdon's full-text patch vs vdelta

["Brian W. Fitzpatrick" <fi...@red-bean.com>]

On Mon, 2004-01-05 at 00:57, Tom Lord wrote:
>     > From: Brandon Ehle <az...@yahoo.com>
> 
>     > Mozilla 1.6b source import
> 
>     > SVN without patch - 158 seconds
>     > SVN with patch - 139 seconds (1.14x)
>     > CVS - 10 seconds (15.80x)
>     > Perforce - failed (exceeded shell parameter limit)
>     > Arch - failed (exceeded shell parameter limit)
> 
> Regarding the `arch' case: excuse me?  Are you testing with an old
> shell version or something?  So far as I know, no version for quite
> some time can possibly generate such an error.  I'm not even
> particularly sure that the shell version would display such problems.
> 
> That you got a similar error for perforce lends credence to the
> hypothesis that you are reporting a bug in your testing procedure
> rather than the tools being tested.

Ease up there Tom.  My understanding of what Brandon said is that he ran
into a limitation of his *shell*, not of arch or perforce.

>        > I've omitted the Perforce and Arch times for the Mozilla
>        > import because I could not do these operations in a single
>        > (recursive) call like SVN and CVS (even though I could get
>        > these times via xargs).

My understanding of what Brandon said here (and *please* correct me if
I'm wrong) is that Subversion allows you to recursively import a whole
directory tree in one command like this:

	$ svn import foodir

but that Arch and Perforce do not have recursive imports so you have to
do something like (again, I'm guessing here):

	$ arch import foodir/file1 foodir/file2 foodir/dir1 ...

So if Brandon used some mechanism (UNIX find, perhaps) to generate all
the filenames in the Mozilla tree and pass them to arch via ARGV, that
he exceeded the limit of ARGC--a limitation of his shell.

Brandon, is that correct?

Anyway, I don't see the need for you to get all defensive here...
nobody's picking on arch.

-Fitz


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

Re: Import performance with David Kimdon's full-text patch vs vdelta

[Brandon Ehle <az...@yahoo.com>]

>>That you got a similar error for perforce lends credence to the
>>hypothesis that you are reporting a bug in your testing procedure
>>rather than the tools being tested.
> 
> 
> Ease up there Tom.  My understanding of what Brandon said is that he ran
> into a limitation of his *shell*, not of arch or perforce.

Yes, the bug lies in my shell.

> 
>>       > I've omitted the Perforce and Arch times for the Mozilla
>>       > import because I could not do these operations in a single
>>       > (recursive) call like SVN and CVS (even though I could get
>>       > these times via xargs).
> 
> 
> My understanding of what Brandon said here (and *please* correct me if
> I'm wrong) is that Subversion allows you to recursively import a whole
> directory tree in one command like this:
> 
> 	$ svn import foodir
> 
> but that Arch and Perforce do not have recursive imports so you have to
> do something like (again, I'm guessing here):
> 
> 	$ arch import foodir/file1 foodir/file2 foodir/dir1 ...
> 
> So if Brandon used some mechanism (UNIX find, perhaps) to generate all
> the filenames in the Mozilla tree and pass them to arch via ARGV, that
> he exceeded the limit of ARGC--a limitation of his shell.
> 

This is exactly the case (only my command is "tla add").  Because I have 
to do the import in multiple calls for Perforce and Arch, I decided not 
to post the results of this test (even though I did obtain such results 
via   "find | xargs tla add" or "find | xargs p4 add") as CVS / SVN 
could have an unfair advantage unless I recompile my shell with a bigger 
limit.


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