There are a couple of warnings from both the string-algo and thread lib doc builds that look like they might be errors: Citation to nonexistent publication abbrev: Citation to nonexistent publication abbrev: Any ideas? Presumably something common to both libs? John.
John Maddock wrote:
There are a couple of warnings from both the string-algo and thread lib doc builds that look like they might be errors:
Citation to nonexistent publication abbrev: Citation to nonexistent publication abbrev:
Any ideas? Presumably something common to both libs?
I am trying to look into this (for thread). However the doc build process does look to me like black magic. Also I could not find anywhere the rationale why building the docs does overwrite files that are under source control. I am trying to compile the doc from BOOST_ROOT/doc. Is this the wrong place? Then the process generates files under bin.v2, which is what I expected. But surprisingly also files under: doc/xml doc/html and under libs/ are generated. Even uglier: some files which are under source control are overwritten! Nevertheless the docs were built, and I logged everything. Nowhere I could find the "Citation to nonexistent publication abbrev:". How did you get this? Please describe step by step. Roland
I am trying to look into this (for thread).
However the doc build process does look to me like black magic. Also I could not find anywhere the rationale why building the docs does overwrite files that are under source control.
I am trying to compile the doc from BOOST_ROOT/doc. Is this the wrong place?
Nope, that's where I built from.
Then the process generates files under bin.v2, which is what I expected. But surprisingly also files under:
doc/xml doc/html and under libs/
are generated. Even uglier: some files which are under source control are overwritten!
Yep, we have "placeholders" that redirect to the nightly doc build, but when the release is put together these get overwritten with the "true" versions.
Nevertheless the docs were built, and I logged everything. Nowhere I could find the "Citation to nonexistent publication abbrev:".
How did you get this?
Just as you did: built from /doc under win32 on the release branch. I've just done another cvs update so I'll try again. John.
John Maddock wrote:
Yep, we have "placeholders" that redirect to the nightly doc build, but when the release is put together these get overwritten with the "true" versions.
This however makes the process somewhat obscure during authoring however...
Just as you did: built from /doc under win32 on the release branch. I've just done another cvs update so I'll try again.
What I found so far: For thread there is an entity.xml file which is beeing used for "short hand" citations too. I just tried to check by hand for missing ones. Nope :-( Is it possible that I do not get the errors, since I am trying the build from linux? Perhaps the tools are different? Btw.: There are so much errors or messages, looking like errors, that I simply wouldn't see the one you are pointing at when not trying to specifically filter for it. To me the whole doc build process looks surprising anyway. The surprise is, that I end up having some readable docs despite the long warning and "error" listsings. But this might be specific to linux. If so, perhaps an indication, that the doc build process doesn't port easily? Btw.: Did you ever try to build bjam pdf for the whole library? Roland
Roland Schwarz wrote:
What I found so far:
For thread there is an entity.xml file which is beeing used for "short hand" citations too. I just tried to check by hand for missing ones. Nope :-( Is it possible that I do not get the errors, since I am trying the build from linux? Perhaps the tools are different?
Btw.: There are so much errors or messages, looking like errors, that I simply wouldn't see the one you are pointing at when not trying to specifically filter for it.
Tell me about it :-(
To me the whole doc build process looks surprising anyway. The surprise is, that I end up having some readable docs despite the long warning and "error" listsings. But this might be specific to linux. If so, perhaps an indication, that the doc build process doesn't port easily?
The main problems are messages relating to missing linkends: those indicate broken links.
Btw.: Did you ever try to build bjam pdf for the whole library?
No! Runs screaming to the hills.... John.
Nevertheless the docs were built, and I logged everything. Nowhere I could find the "Citation to nonexistent publication abbrev:".
How did you get this?
Looking into this some more, I think I only saw this because I'm using Bleeding-edge stylesheets, there is some discussion here http://www.oasis-open.org/archives/docbook-apps/200609/msg00021.html regarding whether the behaviour is correct or not. The warning is issued if you have a citation to a non-existant entity, but unhelpfully doesn't tell you which one :-( My gut feeling is that the warning might indicate a typo in the XML but I haven't tracked anything down yet... trying again with the very latest stylesheets. John.
Roland: I've decided this must be a red-herring, I've tracked down the <citation> entries in the docbook XML that generate the warning, and they *do* lead to valid links in the html, so I think we can ignore this one, although I admit I don't really know what does and does not cause the warning. Apologies for the noise! John.
participants (2)
-
John Maddock -
Roland Schwarz