On 8/24/06, Jean T. Anderson <[EMAIL PROTECTED]> wrote:
<snip errors/>
These warnings and errors aren't very helpful.
Plus it's a case of double indirection. The error you see is not from
processing any specific dita file, but from applying a substylesheet
to the merged FO file.
Oddly enough 'ant all'
seems to provide what looks like helpful details about specific files.
/home/jta/DB-DERBY-SVN/derby/docs/trunk/out/reftemp/rreffuncxmlexists.html
[xslt]
file:/home/jta/DB-DERBY-SVN/derby/docs/trunk/DITA-OT1.1.2.1/xsl/common/output-message.xsl:43:16:
Warning! ------------------------------------------------------------------
[xslt] IDXS015E Error: (File =
/home/jta/DB-DERBY-SVN/derby/docs/trunk/src/ref/rreffuncxmlexists.dita,
Element = xref:1)
[xslt] Unknown file extension in href: http://www.w3.org/TR/xpath
[xslt] If this is a link to a non-DITA resource, set the format
attribute to match the resource (for example, 'txt', 'pdf', or 'html').
[xslt] If it's a link to a DITA resource, the file extension must
be .dita or .xml .
It's complaining about:
<xref href="http://www.w3.org/TR/xpath";>http://www.w3.org/TR/xpath</xref>
which should be either:
<xref href="http://www.w3.org/TR/xpath";
format="html">http://www.w3.org/TR/xpath</xref>
or:
<xref href="http://www.w3.org/TR/xpath.html";>http://www.w3.org/TR/xpath</xref>
The first form is probably preferable. Not sure if that will fix the
overall problem, though.
I hope this is helpful for isolating the problem. I think the key is to
run at least 'ant pdf', 'ant html' and 'ant monohtml'. 'ant all' is
best, especially for debugging it would seem, but I appreciate how long
that takes.
You should still be able to see the problem when running any of the
other ref builds: html.ref, monohtml.ref, or pdf.ref. Please let me
know if that is the case. Also, I think what is needed are individual
targets for the manuals, e.g. 'ant devguide' builds all three formats
for the devguide. Not sure why noone asked for it yet, but I'll add
that.
It does take quite a while to run 'ant all' and the problem is
twofold: one is that currently the PDFs are being generated twice,
and the second is that the Reference Manual requires so much memory to
build that you will most likely end up using virtual memory unless
your machine has over a GB of memory. I will be checking in a fix for
the first problem shortly, but there's not much I can do about the
second one. On my main machine, it takes an hour to run 'ant all'
without the patch, and only about 30 minutes to run with the patch.
The patch simply eliminates the 'all' target's dependency on the 'pdf'
target, since all the dependents of the 'pdf' target will have already
been called by the various monohtml targets.
There's still room for improvement, but I'm not sure how to tackle it.
It would be great if the DITA preprocessing steps before building the
XHTML/FO could be shared between those targets so that they don't need
to occur twice. That would probably bring the time for 'ant all' in
the docs down to the sub-ten-minute range. The fix might need to be in
the DITA toolkit, unless I can find a way to chain the FO build to the
XHTML without all of the temp files being removed. Anyway, something
for a later date...
andrew
