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

Re: path.c:113: svn_path_join: Assertion `is_canonical (base, blen)' failed.

From: Greg Hudson <ghudson_at_MIT.EDU>
Date: 2004-08-26 20:58:21 CEST

On Thu, 2004-08-26 at 14:43, Ben Reser wrote:
> Well it's not canonical if you're treating it as a local path. But it
> is canonical as a URI. Which is what I realized a little bit further
> down in the email. The problem is we have no way to say to
> svn_path_canonicalize() "We never accepts URLs. Canonicalize as though
> everything were a local path."

I agree, that should be fixed. There should be an entirely separate set
of functions to operate on URIs.

> IMHO canonical path is defined by the output of svn_path_canonicalize().

Only if what you passed to svn_path_canonicalize() is a path. If it's a
URI, then what you have is a canonicalized URI.

> But ignoring who's right or wrong let's look at the implications of your
> view here. If svn_path_dirname() does not accept URLs that means every
> app that uses it *MUST* call svn_path_is_url() prior to calling it to
> ensure they won't get an assert.

Yes, because svn_path_canonicalize() has a broken contract. But that's
good for usability anyway; if a function only operates on paths, not
URIs, then it should produce an intelligent error if handed a URI.

> Additionally, svn_path_is_url is documented as:
> /** Return @c TRUE iff @a path looks like a valid URL, @c FALSE
> otherwise. */
>
> Uhh so paths can be URLs!?!? This totally doesn't agree with what you
> seem to be saying. Neither does the naming of svn_path_is_url().
> In the end apps are going to have to be writing:

That's a poorly-named function and a poorly-named parameter, perhaps
because there's no simple word for path-or-URL. But it's not really a
compelling argument for the idea that URLs are always valid arguments to
the svn_path functions. And there are other arguments against; for
instance, the svn_path_join docstring says:

 * Note that the contents of @a base are not examined, so it is possible to
 * use this function for constructing URLs, or for relative URLs or
 * repository paths.

(Of course, that turns out to be false; svn_path_join yields an
assertion failure if base is not canonical.) Most of the svn_path
docstrings are quite explicit when they accept URLs.

> if (svn_path_is_url (path))
> {
> // their own path splitting stuff for URLs that works properly
> }

Do we ever deliberately call svn_path_dirname on a URI currently? Is
there any reason to do so? It seems like a layering violation to me.

Incidentally, if you call svn_path_canonical() on "http://hostname/", it
looks like the result will be http://hostname/, which will fail the
canonical check. Does that seem correct to you?

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu Aug 26 20:58:47 2004

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