fromHtml method blows pdf graphics stack

[mcurland]

The dispose method in the from_html plugin does two things. First, you pop the graphics stack operator pushed in the init method (the Q in dispose pops the q in init). Second, you return an object with the current x/y positions.

The problem is that dispose is called multiple times in the process method, once by the return and the other just above it inside the 'checkForFooter.call' function via callback(r.dispose()).

You either need to not use the dispose method to get the current position for the callback, or protect the write("Q") call so it only fires once. Feedback on the intent would be nice.

As far as I can tell this breaks the graphics stack for all calls to fromHtml. The Chrome and FF viewers handle the bogus pdf gracefully, but adobe reader does not. Any additional elements on the same page simply do not render. Any pdf validator (such as http://www.pdf-tools.com/pdf/validate-pdfa-online.aspx) will also catch the stack error.

[MrRio]

Hmm interesting. Thanks very much for the ticket. Could you create a pull request?—
Sent on the run. Apologies for brevity.

On Sat, Aug 30, 2014 at 9:31 PM, mcurland notifications@github.com
wrote:

The dispose method in the from_html plugin does two things. First, you pop the graphics stack operator pushed in the init method (the Q in dispose pops the q in init). Second, you return an object with the current x/y positions.
The problem is that dispose is called multiple times in the process method, once by the return and the other just above it inside the 'checkForFooter.call' function via callback(r.dispose()).
You either need to not use the dispose method to get the current position for the callback, or protect the write("Q") call so it only fires once. Feedback on the intent would be nice.

As far as I can tell this breaks the graphics stack for all calls to fromHtml. The Chrome and FF viewers handle the bogus pdf gracefully, but adobe reader does not. Any additional elements on the same page simply do not render. Any pdf validator (such as http://www.pdf-tools.com/pdf/validate-pdfa-online.aspx) will also catch the stack error.

Reply to this email directly or view it on GitHub:
#348

[diegocr]

@mcurland, you rock :)

Apparently that's the reason for the recently found bug discussed at #355 and thanks to your mention of checkForFooter usage we should guess it was introduced on #260... if so, surprising we didn't found the bug earlier, and somewhat confusing since we tested that and IIRC we didn't noticed any such issue with missing images - I wonder if perhaps some of the latest commits made the bug worse and more quickly noticeable, dunno.

@woolfg what do you think?

[diegocr]

Er, just did a quick look at the code and it's all i needed to properly understand the issue... :$

Well, it should be fixed now on commit b675300

@mcurland Can you please verify? I think that's all, getting rid on the r.dispose() call on process() return.

@woolfg sorry for bothering you!

[woolfg]

thx for investigation, seems that r.dispose() is called twice, don't know why we didn't recognized this problem...would be great if @mcurland may verify the newest commit.