I do hope I have this right: I don't think this one is down
to my rendering environment.
As I think of it, there are three files created in RENDERTMP
lfs-html2.xml
lfs-full.xml
lfs-html.xml
resulting from this pipeline (simplfied) in the Makefile
xsltproc index.xml --output lfs-html2.xml
xmllint lfs-html2.xml -o lfs-full.xml
xsltproc lfs-full.xml --output lfs-html.xml
and then the Chunked book gets created by an
xsltproc lfs-html.xml
The noChunks book, however, gets created by an
xsltproc lfs-html2.xml --output LFS-BOOK.html
Note that the "source" of the noChunks book is not the
same as that of the Chunked book.
Note also that validate target in the Makefile does a
./aux-file-data.sh $(RENDERTMP)/lfs-full.xml
and that that aux-file-data.sh does some substituion
inside its file argument, as follows
sed -i -e s/BOOTSCRIPTS-SIZE/$bootsize/ \
-e s/BOOTSCRIPTS-INSTALL-KB/$bootinstallsize/ \
-e s/BOOTSCRIPTS-MD5SUM/$bootmd5/ $FILE
What this seems to mean is that the noChunks book doesn't
"see" any changes made by aux-file-data.sh and so ends up
rendering, for example, in Chapter 3's package listing
LFS-Bootscripts (20191031) - BOOTSCRIPTS-SIZE KB:
rather than
LFS-Bootscripts (20191031) - 32 KB:
It also means that the noChunkc book isn't being created
from the xmllint-ed source either.
As I'm not sure why the noChunks book gets rendered from the
very first source in the temporary chain, it's not clear which
is the correct way to address this.
FWIW, the ways I've thought of so far are:
1) Render the noChunks book from the second or third sources,
so lfs-full.xml or lfs-html.xml, in the temporary chain.
2) Run the aux-file-data.sh against the first source
in the temporary chain, lfs-html2.xml, as well as,
or before, running it on lfs-full.xml
Assuming there's no reason why the noChunks book shouldn't
be created from the xmllint-ed source, (1) looks the cleaner
solution.
--
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
