Ever helpful. Thanks a ton.
The italics-outside-the-box is not an issue here, more a curiosity.
I'll try other fonts to see what affect that has.
Now that I understand the value of the line numbers, I make good use of
them by running from the commandline and trapping the output. (I must
turn down the logger though because fop1.1rc1-VH is generating 700Mb log
files!)
I've drastically reduced the overflow, at least for the parts which I
can control. I had been using fixed height in certain components and
these were involved in the vast majority of the overflow. Since we are
at a point in our process where clipping the content is no longer
desirable I have removed the height restrictions (and overflow=hidden)
and am much cleaner now.
In the parallel thread, "Spurious indentation with (small) text
lose[sic]" you can find example fo with no over /except /for the
region-body. Unfortunately the affected region-body elements are not
supposed to have any text. My guess is that overflow is the fragment of
data being lost. I suspect also this same situation was in play with
your patch. Since I have removed the bulk of the overflow fop-1.0 still
produces page-number stutters and that may still be as you point out due
to the overflow - again on region-body elements which have zero sized.
My bet is that both my issues are tied to the same flaw, with apparently
separate manifestations.
Welcome home :)
rjs
On 08/09/2012 01:18 PM, Vincent Hennebert wrote:
Fix committed in rev. 1371386:
http://svn.apache.org/viewvc?rev=1371386&view=rev
See also Bugzilla 53688:
https://issues.apache.org/bugzilla/show_bug.cgi?id=53688
As to your other questions:
On 20/07/12 22:49, Rob Sargent wrote:
Saint Vincent, you deliver us again. The patch works on the original fo! (One
vote from here for committing the patch)
About the overflow:
* All the incidents I've looked at are in the captions under the
header or in the page header. Would overflow in these static
regions the cause of the page stutter?
No. The stutter was caused by an overflow in the region-body.
* Said captions are rendered in font-style="italic". Placing a border
around the image captions I find that the slant forces the bottom of
the first letter of each line outside the border. Is this not a bug?
For instance if I had another element abutting the caption, the
first letter of each line would overlap that abutting element.
Difficult to tell without having the font. There might be a bug in
reading the font metrics, but it’s more probable that the glyph’s
outline sticks out of the bounding box. In other words, that the glyph
is bigger than it pretends to be. A solution could be to add some
padding to the element that has the border.
* The border on the caption does get clipped (2/3 of the time the
sides and the bottom does not appear in the printout), but I'm at a
loss to see why. There's plenty of room according to what is shown
if I add a background colour to the caption text block and a
different colour for the static region.
I’m not sure what you are referring to. You may want to mention this
issue in a new thread (if you haven’t already done so).
* Btw, when running the application in IntelliJ all the overflow
messages show as being at position(-1, -1), so that was rather
confusing.
This is probably because you pipe the result of the XSLT straight to
FOP, without saving it to a file. In that case there is no information
about line and column. If you save it to a file and format it nicely,
then you will get useful information about the locations of problems.
HTH,
Vincent
And another definitions of overflow: the log of my full-chapter fo was 450M!
On 07/20/2012 04:37 AM, Vincent Hennebert wrote:
I've run out of the time I can allocate to this issue I'm afraid. I have
a fix but it needs to be cleaned up and tested.
The best way to avoid the problem is to fix all the overflow warnings
that you are getting. You probably want to fix them anyway since it may
lead to a better output.
If constraints prevent you from fixing them all, then you can try the
attached patch against the latest trunk. It should be fairly safe in the
sense that it doesn't modify the layout engine's behaviour, only the
overflow reporting mechanism. That said, it's untested and unsupported
:-)
In the future, one thing that would be of great help when you submit FO
files would be to replace all the custom fonts with the default ones:
serif, sans-serif, monospace. That will avoid a lot of noise in the
warnings, but most of all allow us to reproduce the exact same layout as
you have.
Thanks,
Vincent
On 19/07/12 22:09, Rob Sargent wrote:
A prince you are!
I do have many of those messages, some are very regular (same 7200pt overflow)
and must be from a repeated element. Others are more varied are hence more
likely to be in the flow perhaps? Since as you point out they are not page
numbers, is there anything I can do, e.g. command line params, debug levels to
help me pin-point the errant content?
Counting this one as under control, I still have one more issue for which I
think there is an page break problem. Still working on a small repeatable
offender.
Cheers,
rjs
On 07/19/2012 02:26 PM, Vincent Hennebert wrote:
Hi Rob,
Thanks for your efforts to reduce the size of the FO file, this is
really appreciated.
I believe I found the problem, but need to investigate a bit more. In
the meantime, try to modify the content so that it doesn't overflow the
page. Watch for messages saying 'Content overflows the viewport of the
fo:region-body on page n'. Except that 'n' is not the page number... but
the column number!
If you can avoid page overflows, you should avoid triggering the issue
and get appropriate page numbers.
HTH,
Vincent
On 19/07/12 20:33, Rob Sargent wrote:
I have whittled the fo down further (shorter-fop.xml, now 35%) of original,
but this is as small as I can go to reproduce the exact 3-page stutter.
I can
remover one more page and get a single stutter (shortest-fop.xml)
I pray this will be enough.
Thanks a bunch.
On 07/19/2012 12:51 PM, Rob Sargent wrote:
I truly regret the large fo; I know it hampers things at debug time. I
had
to find a particular starting point such that the content was still aligned
with the original placements. I'll try to trim more pages from after the
page numbering problem.
rjs
On 07/19/2012 01:31 AM, Pascal Sancho wrote:
Hi,
To be efficient, a short fo helping to reproduce the issue should be
attached to the bug entry, so opening a bug rather than ask on this
list doesn't help anymore if the material remains the same.
I (or somebody else) have to look closer into the fo you attached
here, then perhaps I (or somebody else) will, or will not, find a bug
causing this issue. Since the use case is quite big, that will take
some extra time :/
2012/7/18 Rob Sargent <rsarg...@xmission.com>:
My apologies, but I've been unable to shrink this to a small test. I'm
forced to send a rather large fo, but that's my level of desperation. Or
should I open a bug report?
rjs
On 07/12/2012 09:20 AM, Rob Sargent wrote:
I will do my best, but I see nothing in the generated fo that strikes
me as
suspicious.
See other post on not getting identical output from commandline, which is
hampering generating reliable tests.
Anecdote: prior to the 3 pages of stutter there are three pages
(correctly
numbered) which have no text (entirely region-before), so this may
also be
related to issue
If you're anxious for an example, I can certainly post the 40-page
chapter
fo but the last one made Vincent um, er, grumpy :).
rjs
On 07/12/2012 02:35 AM, Pascal Sancho wrote:
Hi,
One needs to see what there is in the XSL-FO (XML + XSLT transfo
result) fed to FOP.
Can you have a look on it, or provide a minimal version of it, so one
can reproduce the described issue?
2012/7/11 Rob Sargent <rsarg...@xmission.com>:
In the middle of my document that page numbering goes 19, 20,
24,24,24,24,
25. Further more the table-of-contents generated by
fo:page-number-citation
uses the bogus 24s for elements on actual page 21,23.
Anyone seen this before? Hints on rooting out the cause?
FOP-1.0
rjs
---------------------------------------------------------------------
To unsubscribe, e-mail: fop-users-unsubscr...@xmlgraphics.apache.org
For additional commands, e-mail: fop-users-h...@xmlgraphics.apache.org
