Re: Bug in M4

[Stepan Kasal]

Hello,

[again, I'm moving the discussion to bug-m4]

in the bug report below, you say that the text saved by m4wrap should
not be processed at the end of the input when the m4 run is to be frozen.
The text should be saved to the frozen file instead.

I read the node `Frozen files' of the GNU m4 manual, and I think you are
right.  The diversions are already handled like this, and the text saved
by m4wrap should be handled analogically.

I guess this change could be introduced in the next version of m4, as I
hope that people don't usually use m4wrap to get text executed just
before a frozen file is created.

I'm just a member of the bug-m4 list, so my word cannot be taken as a
final decision.  If the maintaners/developers agreed, another question
would be who has time to implement this.  Do you have capacities to
write and donate a patch?

Thank you very much for reporting this problem,
        Stepan Kasal

[The test file from the original bug report is attached to this mail, too.]

On Sun, Jan 09, 2005 at 08:00:44PM +0100, Sam Lauber wrote:
> I am using M4 1.4.2.
> I wrote an example of m4wrap which did the following:
> 
> define macros
> initalize by m4wrapping to an action to remove a temp file
> 
> However, I ran m4 -R on it, and it said
> 
> (my initalizing message)
> (my cleaning-up message)
> 
> I expected
> 
> (my initalizing message)
> 
> I ran m4 -F on the generated file, and pressed EOF. I got
> nothing. I expected
> 
> (my cleaning-up message)
> 
> This seriously reduces the usefulness of m4wrap, since large
> applications have to establish cleanup actions, and large
> applications typically are in frozen files.
> My input file was attached.
> 
> Samuel Lauber