Re: [pdf-devel] [pdf_fsys_item_p] function failed with an existing file

[antoine . humbert]

----- "Aleksander Morgado" <address@hidden> a écrit :

> Hi Antoine,
> 
> >
> > when using the pdf_fsys_item_p function with an exsiting file, the
> function return PDF_FALSE.
> 
> Not sure if that is because the pdf_fsys_item_p function failed or
> due
> to the problems in managing he pdf_text_t strings... see below.
> 
> >
> > path to file has been initialized with
> > pdfHostEncoding = pdf_text_get_host_encoding();
> > pdf_text_from_unicode("/path/to/test.pdf", 17, pdfHostEncoding,
> &text);
> >
> 
> You can't use a host encoding in pdf_text_from_unicode(). Try to use
> the following and let me know if the problem in the pdf_fsys_item_p
> continues:
> pdf_text_from_unicode("/path/to/test.pdf", 17, PDF_TEXT_UTF8, &text);
> 

Sorry, I misspelled the name of function I uses : it was pdf_text_new_from_host.

I also tried pdf_text_from_unicode("/path/to/test.pdf", 17, PDF_TEXT_UTF8, 
&text);

but the result is the same


> > a call to pdf_fsys_item_p(NULL, test) return PDF_FALSE however
> /path/to/test.pdf exists
> >
> > an strace on the program call show that the access() function is
> called with following parameters :
> > access("/path/to/test.pdf \251\4\10q", R_OK)     = -1 ENOENT (No
> such file or directory)
> >
> > It seems that string parameter given to access is incorrect. It
> looks like the string in the pdf_text_t struct is not NULL
> terminated.
> 
> Inside the pdf_text_t, the string is stored as UTF-32HE (either LE or
> BE), without any last NUL bytes. Anyway, that data can't be used in
> any system call, as it must be converted before to host-encoding
> (posix systems) or to UTF-16LE (for windows).

So, it seems that the pdf_fsys_item_p function, which uses the access system 
call,
has to convert the string in the pdf_text_t before calling the access system 
call.

> 
> >
> > Also, when getting pdf_text_t string to output the result, like this
> :
> > pdf_text_get_host(&pdfString, &pdfStringSize, inputFile,
> pdfHostEncoding);
> >
> > a printf("%d - %s", pdfStringSize, pdfString) gives
> > 17 - /path/to/test.pdf \251\4\10q
> >
> > I have to do pdfString[pdfStringSize] = 0 before calling the printf
> function to have the correct output :
> > 17 - /path/to/test.pdf
> 
> Yes, that's right. Returned string doesn't have last NUL byte. You
> are
> limited in the way that you can't take the output of
> pdf_text_get_host() and use it in printf() call. You can't do it for
> example with a UTF-32 string, as having NUL bytes is more than common
> in that encoding, and printf() would detect end-of-string there. If
> you want to have a string so that you can printf it, try to use
> pdf_text_get_unicode() function requesting for UTF-8 strings. Also,
> if
> you use the PDF_TEXT_UNICODE_WITH_NUL_SUFFIX option in this last
> function, the last NUL byte will be appended to the string, so that
> you can use it in printf for example.
> 
> Cheers,
> -Aleksander