On 04.09.2012 21:07, Branko Čibej wrote:
On 04.09.2012 20:42, Stefan Küng wrote:
Seems I got confused here:
What we have to pass to almost each svn API is the config hash for the
svn_client_ctx_t struct. That's what I need a copy of.
The svn_config_* functions work with svn_config_t object. The tmp_key
and tmp_value fields in those object are in fact the ones that prevent
you from reading the same svn_config_t from multiple threads without
synchronization.
And as Bert already noticed: apr_hash_copy only does a shallow copy,
not a deep copy. Seems I have to iterate over the hash and copy every
item separately.
That is correct. You'll also have to duplicate the svn_config_t
structure itself, I think that essentially means doing an
svn_config_create and copying over "x_values" and
"section_names_case_sensitive"; and, of course, doing the deep copy of
"sections".
If you do the deep copy properly -- i.e., copying the keys and values
into the new config_t's pool as well as the hashes -- then you'll save
yourself a bit of trouble with pool lifetimes, too.
I'll do that first in TSVN, then do some tests.
After that I can use that code to create an svn_config_dup() API if
still required.
Why not just write svn_config_dup in libsvn_subr/config.c? Surely
there's no reason to keep this functionality out of the core libraries.
This will be a new API which won't get backported.
But in TSVN we have the problem now and need to fix it.
So I'll have to implement it there as well, at least until 1.8 is out.
Stefan
--
___
oo // \\ "De Chelonian Mobile"
(_,\/ \_/ \ TortoiseSVN
\ \_/_\_/> The coolest Interface to (Sub)Version Control
/_/ \_\ http://tortoisesvn.net
