Hi
I want to create a null pdf surface to be used for calculating metrics without writing a file. See:
http://cairographics.org/manual/cairo-pdf-surface.html#cairo-pdf-surface-cre...
But, with something like:
(defmacro with-pdf-file ((filename width height) &body body) "Execute the body with context bound to a newly created pdf file, and close it after executing body." `(let* ((*context* (create-pdf-context ,filename ,width ,height))) (unwind-protect (progn ,@body) (destroy *context*))))
(with-pdf-file ((cffi:null-pointer) 500 500) (move-to 100 100) (line-to 200 200) (stroke))
I get (a bunch of) errors spit out:
WARNING: function returned with status WRITE-ERROR.
So I guess it's not working properly. Is there a different way I could try specifying that "NULL" gets passed into the cairo function perhaps? Or is this perhaps just a bug in cairo itself?
scott
On Sun, Jan 04, 2009 at 07:40:40PM -0800, Scott Graham wrote:
Hi
I want to create a null pdf surface to be used for calculating metrics without writing a file. See:
http://cairographics.org/manual/cairo-pdf-surface.html#cairo-pdf-surface-cre...
But, with something like:
(defmacro with-pdf-file ((filename width height) &body body) "Execute the body with context bound to a newly created pdf file, and close it after executing body." `(let* ((*context* (create-pdf-context ,filename ,width ,height))) (unwind-protect (progn ,@body) (destroy *context*))))
(with-pdf-file ((cffi:null-pointer) 500 500) (move-to 100 100) (line-to 200 200) (stroke))
I get (a bunch of) errors spit out:
WARNING: function returned with status WRITE-ERROR.
So I guess it's not working properly. Is there a different way I could try specifying that "NULL" gets passed into the cairo function perhaps? Or is this perhaps just a bug in cairo itself?
Hi Scott,
I am at a conference and I don't have time to investigate this in detail at the moment, but I suspect that the error arises from how CFFI handles :string types. When you use a string type, CFFI allocates a memory area, and copies the string to that before calling the function with the resulting pointer. My guess is that this chokes on NULL.
What I would do is ask on the CFFI list whether there is a clean way to pass a NULL pointer to a CFFI function which expects a string argument. If not (which is my guess), then we have to do the allocation and conversion manually, and call a function with type char*. See Section 4.8 of the CFFI manual on this issue.
If you have time on your hands and know how to program C, you can experiment with the following: write a C function
void test(char* string) { /* open a file with a fixed name, eg /tmp/test, write string into it, close */ ... }
define this in CFFI with the argument of type :string, and see what is written to the file when you call this with cffi:null-pointer.
Anyhow, please ask the smart people who designed CFFI. If possible, please CC me.
Thanks,
Tamas
Ok, thanks. I'll try to dig into it to it a little so I can ask a coherent cffi question.
For now I've opened "/dev/null" as the file. It's probably less efficient, but works at least.
scott
On Sun, Jan 4, 2009 at 8:00 PM, Tamas K Papp tpapp@princeton.edu wrote:
On Sun, Jan 04, 2009 at 07:40:40PM -0800, Scott Graham wrote:
Hi
I want to create a null pdf surface to be used for calculating metrics without writing a file. See:
http://cairographics.org/manual/cairo-pdf-surface.html#cairo-pdf-surface-cre...
But, with something like:
(defmacro with-pdf-file ((filename width height) &body body) "Execute the body with context bound to a newly created pdf file, and close it after executing body." `(let* ((*context* (create-pdf-context ,filename ,width ,height))) (unwind-protect (progn ,@body) (destroy *context*))))
(with-pdf-file ((cffi:null-pointer) 500 500) (move-to 100 100) (line-to 200 200) (stroke))
I get (a bunch of) errors spit out:
WARNING: function returned with status WRITE-ERROR.
So I guess it's not working properly. Is there a different way I could try specifying that "NULL" gets passed into the cairo function perhaps? Or is this perhaps just a bug in cairo itself?
Hi Scott,
I am at a conference and I don't have time to investigate this in detail at the moment, but I suspect that the error arises from how CFFI handles :string types. When you use a string type, CFFI allocates a memory area, and copies the string to that before calling the function with the resulting pointer. My guess is that this chokes on NULL.
What I would do is ask on the CFFI list whether there is a clean way to pass a NULL pointer to a CFFI function which expects a string argument. If not (which is my guess), then we have to do the allocation and conversion manually, and call a function with type char*. See Section 4.8 of the CFFI manual on this issue.
If you have time on your hands and know how to program C, you can experiment with the following: write a C function
void test(char* string) { /* open a file with a fixed name, eg /tmp/test, write string into it, close */ ... }
define this in CFFI with the argument of type :string, and see what is written to the file when you call this with cffi:null-pointer.
Anyhow, please ask the smart people who designed CFFI. If possible, please CC me.
Thanks,
Tamas
cl-cairo2-devel@common-lisp.net