Re: r.i.p. MANDOCERR_SYNTLINESCOPE

[Kristaps Dzonsons <kristaps@bsd.lv>]

On 01/11/2011 22:54, Ingo Schwarze wrote:
> Hi,
>
> here is a long-awaited patch to fix one of the last needless
> fatal errors that sometimes haunt us in real-world manuals:
> When breaking a man(7) BLINE scope, do not error out,
> but instead just delete the clobbered macro from the AST.
>
> This fixes rendition(4) in Xenocara.
>
> I see no regressions.
>
>
> The patch is mildly tricky.
>
> In man_validate, it is too late to fix this.
> Consider:
>
> .SH FOO
> lots of stuff
> .TP
> .SH BAR
>
> That produces this AST:
>
>    .SH BLOCK
>      .SH HEAD: FOO
>      .SH BODY
>        lots of stuff
>        .TP BLOCK
>          .TP HEAD (empty, and BODY is missing, ouch)
>    .SH BLOCK
>      .SH HEAD: BAR
>
> The current check_bline is called while parsing the BAR BLOCK.
> So m->last is already pointing to BAR, and there is no reasonable
> way to get back to the utterly broken, incomplete TP block:
> prev ->  child ->  next ->  child, and iterate the next pointers
> until the last?  Certainly not.
>
> So the interrupted BLINE block has to be killed when starting
> the interrupting macro parse sequence in man_pmacro, just like
> it is done for interrupted ELINE macros.
>
> That gets rid of lots of hooks in man_validate.c, but requires
> a new parse flag for those macros losing the check_bline hook.

Ingo,

Yes, I like this---please commit it.

Thanks!

Kristaps

--
 To unsubscribe send an email to tech+unsubscribe@mdocml.bsd.lv