On 08/27/2018 02:13 PM, Philip McGrath wrote:
I am hoping for some help debugging a problem I'm having writing FFI
bindings for libxml2.
I am trying to use the function `xmlValidateDtd`, which (predictably)
validates an XML document against a DTD. To support error reporting, the
first argument to the function is a pointer to an `xmlValidCtxt` struct
(documented at http://xmlsoft.org/html/libxml-valid.html#xmlValidCtxt,
though note that the last two fields are listed twice because whatever
extracted the docs didn't understand a processor directive). My wrapper
code needs to allocate the `xmlValidCtxt` (via `xmlNewValidCtxt`), set
the first field to a `FILE*` for writing obtained via `fopen`, set the
second and third fields to pointers to `fprintf`, and then call
`xmlValidateDtd` with the instance. I have a little function in C that
does this work:
https://github.com/LiberalArtist/libxml2-ffi/blob/master/myvalidate.c
Of course, I want to write this in Racket, not C, but my attempt to do
this via the FFI
(https://github.com/LiberalArtist/libxml2-ffi/blob/master/segfault.rkt)
causes a segfault. The problem is something in the way I'm initializing
the `xmlValidCtxt` struct. If I run my Racket version with a valid
document, so that the error reporting isn't used, it works just fine.
Likewise, if I don't initialize the first three fields of the
`xmlValidCtxt` struct, instead leaving them as null pointers, the
default error behavior (writing to standard error) works, though it
isn't useful for my purposes. The segfault only happens if I initialize
the fields as described and a validation error actually tries to use the
fields.
I've also confirmed that my C version works. If I compile it to a shared
library and load it with the FFI, it works just as desired with both
valid and invalid documents:
https://github.com/LiberalArtist/libxml2-ffi/blob/master/use-my-so.rkt
I don't see anything different between my attempt to initialize the
`xmlValidCtxt` struct from Racket and the way I'm doing it in C, but
obviously there is some difference. Any debugging suggestions would be
appreciated.
It works for me if I change the definition of `_fprintf-ptr` to this:
(define _fprintf-ptr _fpointer)
The docs say that compared to `_pointer`, `_fpointer` skips a level of
indirection on symbol lookups. (This also might explain a problem I had
in the past getting the right value for a symbol whose value is an array
of pointers to structs. IIRC, I got the pointer to the first struct
instead.)
Ryan
--
You received this message because you are subscribed to the Google Groups "Racket
Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to racket-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
