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

RE: [PATCH] Change error reporting for xlate problems

From: Bert Huijben <bert_at_qqmail.nl>
Date: Sun, 14 Jul 2013 12:38:52 +0200

> -----Original Message-----
> From: Stefan Sperling [mailto:stsp_at_elego.de]
> Sent: zondag 14 juli 2013 11:39
> To: Justin Erenkrantz
> Cc: Daniel Shahaf; Subversion Development
> Subject: Re: [PATCH] Change error reporting for xlate problems
> On Sun, Jul 14, 2013 at 02:06:02AM -0400, Justin Erenkrantz wrote:
> > On Sat, Jul 13, 2013 at 8:39 PM, Daniel Shahaf <danielsh_at_apache.org>
> wrote:
> >
> > > It appears as soon as svn_utf_cstring_to_utf8() is called --- which is
> > > normally during argv parsing. The error happens even if the argv
> > > arguments are all ASCII, which effectively adds a new dependency on
> > > apr_xlate_* even for people who use only ASCII. I assume this new
> > > failure mode for ASCII-only users means we cannot backport this
> > >
> >
> > This now means iconv or apr-iconv are a hard global dependency where
> they
> > haven't been before. I'm not sure that I like that. (If you're going
> > do this patch, remove the if check for EINVAL/ENOTIMPL - it's
> unnecessary.)
> >
> > IIRC, the reason is that this code gets called all the time - we had to
> > silence the ENOTIMPL and EINVAL errors as it really shouldn't be treated
> > a fatal error. So, in this case, we're back to treating it as an
> > irrecoverable error.
> >
> > It's probably better to just check APR_HAS_XLATE and return an error in
> > svnsync if that's 0...and let the string pass through unmodified - which
> > probably fine for svnsync case...or perhaps go a step further and just
> > refuse to build svnsync if iconv isn't supported as we might not be able
> > guarantee the integrity of the sync'd logs. I'm not sure how paranoid
> > need to be here. -- justin
> Requiring iconv for svnsync is a good idea, we need it for the
> --source-prop-encoding option of 'svnsync sync'.

Why do we need this for 'svnsync' in general?

I would guess the only case where we need it would be if the user really
asks us to translate the properties, which in general they shouldn't, as the
properties should be encoded correctly in the original repository.

Note that on Windows we don't use apr-Iconv any more. There is a Windows
specific implementation in our utf subsystem that makes the OS handle the
conversions for us without this dependency.

Received on 2013-07-14 12:39:51 CEST

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