svn_filesize_t (was Re: Issue 1031 (gcc3.3 compile warnings) is out of date)

[Greg Hudson <gh...@MIT.EDU>]

I'm concerned about issue 639, actually.  I don't remember any
discussion of it on the list, and it seems like a rather fundamental
change.  The issue says:

  "However, our sizes should not be related to the operating system
  limits.  We should be targeting a 64-bit file size (and offsets into
  that)."

But there is no justification for that statement.  We have no problem
with file formats, since we never use fixed fields in file formats
(Berkeley DB might use fixed fields internally, but that's not under our
control).  apr_off_t doesn't appear in our svn_client or svn_wc API, and
appears in the svn_fs API in exactly one place (svn_fs_file_length).  So
it seems like issue 639 is going to create a lot of code for no stated
benefit.

(Incidentally, on most 32-bit *BSD systems, off_t is 64 bits and size_t
is 32 bits.  Since some of the CollabNet people are FreeBSD users, I
don't think any warnings on Windows are due to 64-bit off_t being
unusual.)


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

Re: svn_filesize_t (was Re: Issue 1031 (gcc3.3 compile warnings) is out of date)

[Branko Čibej <br...@xbc.nu>]

Greg Hudson wrote:

>I'm concerned about issue 639, actually.  I don't remember any
>discussion of it on the list,
>
But there were such discussions.

> and it seems like a rather fundamental
>change.  The issue says:
>
>  "However, our sizes should not be related to the operating system
>  limits.  We should be targeting a 64-bit file size (and offsets into
>  that)."
>
>But there is no justification for that statement.  We have no problem
>with file formats, since we never use fixed fields in file formats
>(Berkeley DB might use fixed fields internally, but that's not under our
>control).  apr_off_t doesn't appear in our svn_client or svn_wc API,
>
But it appears in svn_dirent_t in svn_types.h. Which means it appears
most everywhere that matters.

> and
>appears in the svn_fs API in exactly one place (svn_fs_file_length).  So
>it seems like issue 639 is going to create a lot of code for no stated
>benefit.
>
>(Incidentally, on most 32-bit *BSD systems, off_t is 64 bits and size_t
>is 32 bits.  Since some of the CollabNet people are FreeBSD users, I
>don't think any warnings on Windows are due to 64-bit off_t being
>unusual.)
>  
>

The point of the fragment you quote is that our API and our schema
should remain the same, regardless of whether apr_off_t is 32- or 64- bit.

-- 
Brane Čibej   <br...@xbc.nu>   http://www.xbc.nu/brane/


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