Re: Bug in current version of Texinfo

[David Kastrup]

Karl Berry <address@hidden> writes:

> First, as for the real point, straightforward index entries with \ in
> the term work, of course.  I maintain several Texinfo manuals about
> (La)TeX stuff, after all.  Can you show me your Texinfo source in
> question?

LilyPond repository on Savannah, ./autogen.sh && make && make doc

Expect make doc to take more than an hour (if you have dual processors,
running CPU_COUNT=3 make -j3 && CPU_COUNT=3 make -j3 doc might be called
for).  PDF files should appear in Documentation/out I think, possibly
Documentation/out-www.  texinfo.tex is in tex/texinfo.tex.  What might
be involved here is that we have several of our own indexing command
definitions and routing through those might break under the current
Texinfo definitions.

I am pretty sure those changes were not gratuitous, and we just recently
had an update to a newer Texinfo version by some worker after which the
documentation master forward-ported this indexing change (which I did
some time ago) in a separate later commit.  I should be surprised if
this action was not prompted by something failing again.

> As for the side issue:
>
>     texinfo.tex is rather full of whitespace errors (mostly trailing spaces
>     at line ends
>
> I don't consider that an error.  If there is any truly semantic
> whitespace error, like the bogus newline (an editing error) you pointed
> out, of course I want to know that.  In general I try to be extremely
> careful about it.
>
> In my experience, the only "nuisance" is from VC or other software that
> cares about something as meaningless as trailing whitespace instead of
> just preserving original text.

Cursor movement in Emacs behaves erratic with trailing whitespace and it
affects search&replace patterns.  Editing operations tend to clean up
depending on the settings of the editor.  As one result, the LilyPond
version of texinfo.tex has garnered a solid dose of gratuitous
whitespace changes that make comparing and merging with upstream more
complex.  Since some of the whitespace errors appear and disappear in
upstream, they are flagged in VC output, cluttering more relevant
information.

> I know it is a widespread convention in GNU now, but I abhor it and
> refuse to kowtow to it.  Sorry.  (It is not just obstinance; in my
> personal environment, that trailing whitespace is actually useful.
> Don't ask.)

The trailing whitespace in texinfo.tex does not seem to follow any
regularity so it is hard to regard it as functional.  And "don't ask"
implies it is functional for a single person.

I tend to think that personal coding style may have better venues of
expression than whitespace.

All the best

-- 
David Kastrup