From: msevior@physics.unimelb.edu.au
Date: Fri Oct 10 2003 - 12:42:47 EDT
> Thanks, Ben.
>
> I've confirmed this on paper. This _could_ be some of
> that "unstable" math that I had previously mentioned.
> Then again, it could be something else entirely.
>
> The printer's DPI (maybe 300?) is quite likely
> different than the screen DPI (96?), thus causing
> niggly effects like this when you do division.
>
> Pat Lam and I are going to work on this ASAP and get
> it into a better state.
>
Great! It could also be the fb_LineBreaker.cpp class. It almost always
works 100% but Ben might have hit on the 0.1% chance it fails.
But I think you're right. My money would be it's caused by rounding errors
at the 1440 dpi you guys chose for layout resolution.
If you want to continue to use integer calculations, you could instead
layout at 7200 dpi. That would allow a 20,000 page document before the
layout units wrap around to negative numbers. So it's OK from that
perspective.
However my gut feeling is that we could comfortablly use double precision
floating point calculations with no noticible performance penalty.
As a real world example, Oliver Schurro reports perfect results on
printing the GUADEC-4 paper.
Cheers
Martin
> Could someone with win32 skillz fix the ruler bug now
> please? It should just be a matter of getting
> 'getGUIFont()' done properly.
>
> Thanks for finding this edge-case. Can you try some
> "real world" examples, maybe, too?
>
> Dom
>
> --- msevior@physics.unimelb.edu.au wrote:
>> -------- Original Message --------
>> Subject: Re: Test printing of the binary and report
>> the results please.
>> From: "Ben North" <ben@redfrontdoor.org>
>> Date: Fri, October 10, 2003 11:47 pm
>> To: <msevior@physics.unimelb.edu.au>
>>
>> Hi Martin,
>>
>> > Thanks for trying the binary. Have you tested
>> printing?
>> > Does it work OK for you?
>> >
>> > Please report good or bad performance to the
>> abiword-dev@abisource.com
>> > list.
>>
>> I'm not subscribed to the list (I follow it on the
>> web via the
>> archive pages), so hope it's OK to reply to you
>> instead.
>>
>> To test printing, I entered the following text as a
>> single paragraph:
>>
>> aaaaaaaaaaaaaaaaaa bbbbbbbbbbbbbbb ncccccccccccccc
>> dddddddddddddddddddd
>> eeeeeeeeeeeeeeeeeeeee
>>
>> Then as a second paragraph, separated by a few
>> linefeeds, entered:
>>
>> aaaaaaaaaaaaaaaaaa bbbbbbbbbbbbbbb mcccccccccccccc
>> dddddddddddddddddddd
>> eeeeeeeeeeeeeeeeeeeee
>>
>> The difference is that the third "word" begins with
>> "n" in the first
>> paragraph and "m" in the second. The attached
>> screenshot shows the
>> results when both paragraphs are set to "justified".
>> The first
>> paragraph fits all on one line; the second breaks
>> onto two because of
>> the slightly higher width of the "m" compared to the
>> "n".
>>
>> When printed out, though, each paragraph appears on
>> just one line. Can't
>> attach a papershot :-)
>>
>> Hope this is helpful,
>>
>> Ben.
>>
>>
>>
>
>> ATTACHMENT part 2 image/png name=abw.png
>
>
>
> __________________________________
> Do you Yahoo!?
> The New Yahoo! Shopping - with improved product search
> http://shopping.yahoo.com
This archive was generated by hypermail 2.1.4 : Fri Oct 10 2003 - 13:04:13 EDT