Re: [Groff] Re: Blank line macro (.blm) evil with `groff -ms'

[Keith MARSHALL]

Werner Lemberg wrote:
>> I would like to report a problem with groff compilation when the
>> "blank line macro" request (.blm) is used.  [...]
>
> Your analysis is wrong.  The problem is completely unrelated to the
> blank line macro.  It is possible to reduce your example to this:
>
>   .ev 1
>   .br
>   .ev
>   .LP
>   x
>
> The call to `.br' invokes groff's output routine.  This in turn
> invokes the address@hidden' macro which is called only once in the whole
> document.  Its function is to initialize the global state of the
> various ms macros, and it assumes that it is in environment 0 (the
> top-level environment).  In the given example, the assumption fails,
> and environment 0 isn't initialized correctly.
>
> I'm not sure whether I should classify the current behaviour as a bug
> or as a feature.  I tend to say it's a feature, and I would like to
> document that the first action which causes vertical movement in the
> output must happen in the top-level environment.  Anything else makes
> the initialization unreliable IMHO.
>
> Comments, please.

Since this hasn't been observed until now, it would appear that the OP
is trying to achieve something rather unusual.  As address@hidden' needs to
be invoked in environment 0, perhaps a warning could be emitted if the
environment has been changed before initialisation occurs?

--- ./groff/1.19.2/tmac/s.tmac.orig      Mon Jul  2 14:28:52 2007
+++ ./groff/1.19.2/tmac/s.tmac   Mon Jul  2 15:05:12 2007
@@ -135,6 +135,10 @@
 .als AB cov*err-not-before-tl
 .de cov*first-page-init
 .rm cov*first-page-init
+.if !'\\n[.ev]'0' \{\
+.      @error environment changed before the first page was started!
+.      @error must be in top level environment when first page is started
+.\}
 address@hidden
 .als RP cov*err-not-after-first-page
 address@hidden

Regards,
Keith.