Re: Please try CVS version [Was: [xmlroff] [ANN] xmlroff 0.3.7]

[Stefan S. <se...@sy...>]

Tony Graham wrote:
> Stefan Seefeld <se...@sy...> writes:
> 
>>Tony Graham wrote:
> 
> ...
> 
>>I get a segmentation fault after lots of 'libfo-CRITICAL' messages. The first
>>one is
>>
>>(process:12015): libfo-CRITICAL **: fo_area_page_size_request:: Cannot split child:: page: 
>>FoAreaPage (0x9e8ddb8 : 2); child: FoAreaViewportReference (0x9e8dcb8 : 1)
>>
>>which I assume your latest work was meant to resolve. Right ?
> 
> 
> I had left the warning in there so I could see when the code was being used.
> It is currently benign.  I should change it to be a warning that is output
> only if the 'overflow' property value is 'error-if-overflow'.
> 
> 
>>The second such message looks like an identical copy of the first, but the third
>>one is strange again:
> 
> 
> That hopefully indicates that you have two pages with content that is too big
> for the page (or you added content to an area that was already too big for the
> page).
> 
> 
>>(process:12015): libfo-CRITICAL **: Expression evaluation error: Expression not completely 
>>evaluated: '/ast.svg'
>>Object path: 
>>/FoTree[1]/root[1]/page-sequence[5]/flow[1]/FoBlockBlock[2]/FoBlockLayout[3]/external-graphic[1]
>>
>>The expression in question seems to stem from an external-graphic attribute
>>src="images/ast.svg". I thought xmlroff is able to deal with such paths. Do I remember
>>wrongly ?
> 
> 
> I think you remember wrongly.  The value of 'src' should be a
> <uri-specification>, which should start with 'url(' and end with ')'.
> 
> You can submit an RFE for xmlroff to support 'src' property values without the
> 'url(...)'.  All it would take would be yet another expression evaluation
> function, since "images/ast.svg" is a plausible expression.
> 
> I don't know that xmlroff supports SVG graphics.  I haven't tried it.
> 
> 
>>The last errors, which appear to have caused the actual segfault, are assertion errors
>>such as:
>>
>>(process:12015): libfo-CRITICAL **: file fo-length.c: line 320 (fo_length_get_value): assertion 
>>`length != NULL' failed
> 
> 
> This may be a byproduct of the external graphic not being created properly.
> It indicates a bug in the error handling previously, since it shouldn't have
> got to the point of trying to process null length datatypes.
> 
> 
>>Let me know if I can help debug this further.
> 
> 
> Please correct your <uri-specification> and, if necessary, try it with a PNG
> instead of SVG image.

Ok, I tuned the xslt stylesheet a bit and now I only get the first two libfo-CRITICAL
errors (warnings). However, xmlroff still segfaults:

0x005a291a in fo_area_size_request_default (child=0x9458540) at fo-area.c:2063

(the rhs 'return_child' is 0x0, and so dereferencing next_part crashes the application)

Let me know if I should mail you the full stack trace.

Regards,
		Stefan