[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: CVS update: subversion/subversion/libsvn_fs clones-table.c dag.c txn-table.c

From: Karl Fogel <kfogel_at_galois.collab.net>
Date: 2001-01-12 16:13:50 CET

Greg Stein <gstein@lyra.org> writes:
> btw, note that JimB agreed that making the string data const was probably
> quite fine. He wasn't so sure about the child/sibling pointers.

Ah, okay. (I had thought JimB had some reason to leave them mutable.)

I'll undo the change and make it go the other way, then.

-K

> [ I think those *can* be, but am happy to defer that question for now ]
>
> Cheers,
> -g
>
> On Fri, Jan 12, 2001 at 02:20:48AM -0800, Greg Stein wrote:
> > Um. What he said.
> >
> > [ as if you couldn't guess that's what I'd say... :-) ]
> >
> > Cheers,
> > -g
> >
> > On Fri, Jan 12, 2001 at 01:53:22AM +0100, Branko Cibej wrote:
> > > kfogel@tigris.org wrote:
> > >
> > > > User: kfogel
> > > > Date: 01/01/11 13:21:08
> > > >
> > > > Modified: subversion/libsvn_fs clones-table.c dag.c txn-table.c
> > > > Log:
> > > > Add casts away from const, to get rid of compiler warnings.
> > > >
> > > > We have const strings being passed into functions that take non-const
> > > > strings, but -- if I understand correctly -- we don't want those
> > > > functions promising the incoming data will never be changed. After
> > > > all, the functions are just storing the data directly (non-copying)
> > > > into skels, and it would be too restrictive if all skels had to
> > > > promise never to modify the data in their strings.
> > > >
> > > > If this is wrong, and skels can make such a promise, then we should do
> > > > this change the other way, adding const to the string parameters of
> > > > svn_fs__str_atom(), svn_fs__parse_skel(), and svn_fs__str_to_dbt().
> > >
> > > Well, skels don't change their contents directly right now, and there's
> > > no API for that. I'm +1e5 for constifying those parameters, and also the
> > > data pointers within the skels (not the child and sibling pointers, of
> > > course).
> > >
> > > Anyway, if we're going to point from skels into constant strings, we'd
> > > better make sure those pointers are to const, or we'll regret it
> > > someday. I think the casts in this case are dangerous. If we can't agree
> > > to constify the pointers, then I'd rather see we leave the warnings
> > > rather than add the casts. At the very least, the poor guy that's gonna
> > > have to fix the segfaults will know where to look.
> > >
> > >
> > > --
> > > Brane Cibej
> > > home: <brane_at_xbc.nu> http://www.xbc.nu/brane/
> > > work: <branko.cibej_at_hermes.si> http://www.hermes-softlab.com/
> > > ACM: <brane_at_acm.org> http://www.acm.org/
> > >
> >
> > --
> > Greg Stein, http://www.lyra.org/
>
> --
> Greg Stein, http://www.lyra.org/
Received on Sat Oct 21 14:36:19 2006

This is an archived mail posted to the Subversion Dev mailing list.