On Sat, Feb 23, 2013 at 12:15:11PM +0200, Eli Zaretskii wrote:
>
> If it's important, I can trace through the C makeinfo and report what
> exactly happens there. Let me know.
I don't think this will be needed, I think the difference that remains
is that in C makeinfo @set and @clear are completly removed from the
output, as is done by the new implementation for @ifclear, for example,
but not for @clear and @set.
> This produces the same output with makeinfo 4.13a. With makeinfo 5.0,
> it mostly works, but there are still a few problems:
>
> . There are many warnings like this:
>
> tramp.texi:481: warning: @set should only appear at a line beginning
> (possibly involving @xxx)
>
> tramp.texi:481: warning: @set should not appear in @file (possibly
> involving @xxx)
>
> tramp.texi:481: warning: @set should not appear in @file (possibly
> involving @xxx)
>
> tramp.texi:1188: warning: @set should not appear in @kbd (possibly
> involving @xxx)
>
> tramp.texi:2345: warning: @set should not appear in @item (possibly
> involving @xxx)
Those have been fixed in svn.
> . There are some errors:
>
> tramp.texi:2457: misplaced { (possibly involving @xxx)
> tramp.texi:2457: misplaced }
> tramp.texi:2457: misplaced {
> tramp.texi:2457: misplaced }
> tramp.texi:2459: too many columns in multitable item (max 1)
> tramp.texi:2460: too many columns in multitable item (max 1)
> tramp.texi:2484: misplaced { (possibly involving @xxx)
> tramp.texi:2484: misplaced }
> tramp.texi:2484: misplaced {
> tramp.texi:2484: misplaced }
Indeed, this error remains, this is linked with
@multitable {@trampfn{telnet, , melancholia.danann.net,}} {@trampfn{telnet, ,
192.168.0.1,}}
and the fact that this is expanded as a multi-line with @set and @clear
which then leads to failures because @multitable arguments should be on
a line.
> . @item @trampfn{, , melancholia, .emacs} and similar constructs
> (see tramp.texi around line 3230) produce this:
>
> '/'
> melancholia:.emacs
>
> instead of this:
>
> `/melancholia:.emacs'
This is a similar issue as above. The svn version, consistently with
leaving the @set in the output, also has a warning:
tramp.texi:2327: warning: @set should not appear in @item (possibly involving
@xxx)
The following is the @multitable issue from above:
> . The text between lines 2453 and 2491, starting with "If you, for
> example, type" and ending with "Now you can choose" produces
> something weird, probably because there are extra @ifset
> conditionals there. Here's the diff against v4.13:
The next issue is completly different, and arises because at some point
there is a @set with an incorrect name, here should be along
@set x$(whoami)
There are other similar issues reported by the svn version:
tramp.texi:978: bad name for @set (possibly involving @xxx)
that correspond to something like
@trampfn{smb,
daniel%BIZARRE, melancholia, /daniel$$/.emacs}
that is expanded as a
@set xdaniel%BIZARRE
The restrictions on @set/@clear flags are indeed new, they are
documented in the manual in @set @value:
A "flag" (aka "variable") is an identifier consisting of
alphanumerics, '-', and '_'. However, it is best to use only letters
and numerals in a flag name, not '-' or '_'--they will work in some
contexts, but not all, due to limitations in TeX.
So, to 'solve' this issue, this specification should be changed.
> . This call:
>
> emacsclient @trampfn{ssh, $(whoami), $(hostname --fqdn), $1}
>
> is converted to "emacsclient /ssh:$(hostname --fqdn):$1", omitting
> the "$(whoami)@" part.
Thus, I think tht most of the issues should be alleviated by removing
completly @set and @clear from the expanded tree, as is done for
@ifclear for example. Should I do that?
The issue that will remain will be the use of non alphanumeric characters
in @set and @clear. Should we change what is specified in the manual?
I'll answer to the customization/documentation issues in a separate
mail.
--
Pat
