Stas Bekman <[EMAIL PROTECTED]> writes:
[...]
> > I still don't see how that's useful here. There's nothing wrong with the
> > data inside apr_table. It's the c2perl
> > translation (newSVpv) that's causing the lossage.
>
> and I don't get where the lossage is :(
OK- I see we're having a disconnect. Let me try to draw a picture
of an apr_table_entry_t that's coming from an apreq-2 parser:
ATTRIBUTE STORAGE (BYTES)
table_entry: key -------------------+
|
v
| f | o | o | \0
val -------------------+
|
|
v
| 12345 | 0 | 8 | b | a | r | \0 | n | o | n | e | \0
^ ^ ^ ^
| | | |
name status size data
Notice that val points to the end of an apreq_value_t struct; what's
important here is that the location of the metadata (ie. the rest of
the apreq_value_t struct) is directly in front of *val.
So what happens when newSVpv copies *val? It ignores the metadata,
so we wind up with unusable (segfaultable) garbage in front of SvPVX:
SvPVX(sv) ----------------------------+
|
|
v
*** apreq_value_t is lost *** | b | a | r | \0
^^^^^^^^^^^^^^^^^^
rest of string is missing
The original metadata isn't lost, it's just inaccessible through the
APR::Table API. But by adding magic to the SV, so that mg->mg_ptr
points at the original C-string pointed at by the apr_table_entry_t's
val pointer, the full metadata can be recovered (via the conversion
macros in apreq.h, e.g. apreq_strtoval()), including the remainder
of the string that newSVpv lopped off.
[...]
> That's cool. But Joe, you don't seem to understand what I'm asking you
> about. I want to see an example of a conversion function where this
> meta data kicks in.
I don't know if I've said enough to answer your question, but
I hope I've explained the problem with newSVpv well enough.
Short of coding up an entire APR::Table subclass, can you be
more specific about what you'd like to see?
> I don't seem to get into your head to see things the same way you do.
Dreadfully laconic, I know. Sorry about that.
My thesis advisor used to say the same thing :-/.
--
Joe Schaefer
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
