Re: use of AC_SUBST

[Jim Meyering]

Bruno Haible wrote:
> Jim Meyering wrote:
>> your patch was already moving the AC_SUBST use,
>> which was technically unnecessarily.
>
> Huh? There is a 'case' statement in gl_GETHRXTIME. If the user had
> set LIB_GETHRXTIME as an environment variable and the first alternative
> of the 'case' statement was chosen, the value propagated into

Yes, making the initialization unconditional is the fix.
Why the "Huh?"  Technically, there's no need to move the AC_SUBST.

> config.status would have been the value of the environment variable,
> not the empty string - leading to build problems. Therefore I moved
> the LIB_GETHRXTIME= initialization. And then I moved the
> AC_SUBST invocation because the code was using the traditional
> coding style explained in the last mail.

My impression is that the traditional style makes it so the initialization
and AC_SUBST *textually* bracket all uses of the variable.  When they
do not (as induced by your patch), it's less valuable, and perhaps even
obtuse, since the other assignment is in a separate autoconf macro that
textually follows the AC_SUBST.

>> If you want to be pedantic, you would not have moved it at all,
>> rather than impose *your* style.
>
> That style was already present in the previous version of the file.
> You cannot say that I imposed it when Paul Eggert coded it that way.