Happy new year and new bug found in m4.

[aikido87]

First of all, thanks for all your work, good people at gnu project.

Second, Happy New Year, also for your families. 

Third, I use the version " GNU m4 1.4o" of m4. Well, now the bug.

I found a "buggy" acting of builtin "undivert" when is used to read a file.

Let's assume that the file "file.txt" has just the word "baby":

 

> define ('xyz' , 'undivert( file.txt)')

> I want you xyz .

=> I want you, baby . 

 

But, if I use "xyz" as a parameter ( of an user defined or a builtin macro), m4 
expands "undivert" before the macro it self, leaving the parameter "void". An 
example with "ifelse":

 

> define (' baby', ' woman')

> I want you xyz .

=> I want you, woman .

>  ifelse (include (file.txt), undivert (file.txt), ' The results of "include" 
> and "undivert" is the same', ' The        

   results is NOT the same').   

 

Now, I expect that simply the string "The results is NOT the same" is expanded, 
since the macro "baby" is now defined.

This is, nevertheless, the output:

 

=> woman The result is NOT the same.

 

If you debug/trace, you can see it acting like:

 

> undivert (file.txt) ''ifelse (include (file.txt), '', . , . )

 

that is, "undivert" seems to be "traslated" before the macro -in this case, the 
macro ifelse-, AND the macro has the second parameter -that one related to 
"undivert"- void.

I tried also with "ifdef" and an user defined macro, and I saw the same acting. 
I would precise that it seems to be indifferent: either I use "undivert" 
directly, or I use it through a macro that calls it -like "xyz"-, the result is 
the same.

On the other hand, if I use "undivert" inside a pure string, no problem at all: 
it acts like I expect.

Fortunately, the builtin "include" seems to have not this problem, either used 
in a pure string or as parameter of other macro.

To finish this bug segnalation, it's my pleasure giving you a little script in 
m4 in wich I did take a challenge: writing a bare "shell" in m4 language (that 
is: emulating "read" for INPUT in a variable, "echo" for OUTPUT, and 
"comm-subst" for the " shell command substitution" -that is the result of a 
command in a variable-.

I give it to you for a reason: please, try to substitute "include(.)" by 
"undivert" (.)" . you'll see the bug.

(That is the way I found it).

The script is in Italian, but I think is quite auto-explicative/easy-to-follow. 
Nevertheless, if you want I can translate it in English (this is not mandatory 
to see the bug).

Last but not least, I would like read your opinion on a question: why GNU did 
not use BSD license and created GPL license ? What's wrong with BSD license ?

You could answer about the m4 bug and the last question at address of this 
e-mail or at this other address:

"vaisarger[AT]yahoo[DOT]it.

 

Thank you!

 

Vittorio Cagnetta, Italy

script_in_m4
Description: Binary data