[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Conditional append

From: J.T. Conklin
Subject: Re: Conditional append
Date: Tue, 11 Jan 2005 10:01:42 -0800
User-agent: Gnus/5.1006 (Gnus v5.10.6) XEmacs/21.4 (Security Through Obscurity, berkeley-unix)

Stepan Kasal <address@hidden> writes:
>> All that cruft could be removed if automake interpreted "+=" of a
>> previously unseen variable as an assignment.
> ``a lot of code''?
> All you need is:
> 1) remember the list of all variables used
>    (To be exact, you don't need the list of variables actually used,
>    any superset will do.)
> 2) output empty assignments to an automake fragment
> 3) `include' that fragment at the beginning of your

"A lot" can be subjective.  It's code that doesn't have to be there if
automake supported += with undefined variables.

MPC doesn't make it easy to remember whether a variable is emitted.
It sucks up *.mpc files and builds lists (source files, header files,
etc.), which it uses to expand *.mpd template that has a minimal set
of control flow and output formatting macros.

Normally MPC also generates a separate file per project.  Since this
would result in 50+ files in some ACE/TAO directories, we merge them
into a single instead of using "include".  Fortunately,
this gives us an opportunity to transform "=" to "+=", but it's done
by parsing the per-project's.

> This doesn't need much code, and you have a which looks
> the same as it would if Automake interpreted += of an unseen
> variable as an assignment.
> But I think that your current code produces nicer's.  If
> a variable is used only on one line, it's more natural if it's an
> assignment.  That matters if there is a danger that a human will try
> to read them.
> So the price is for nicer output, not to overcome Automake
> limitations.

True.  Currently MPC doesn't generate 100% perfect files,
so I have made extra effort to make the output as close as I can to 
what a human would write, since I end up tweaking the output.


J.T. Conklin

reply via email to

[Prev in Thread] Current Thread [Next in Thread]