Build workflow

Russ Allbery rra at stanford.edu
Mon Mar 20 02:34:16 UTC 2006


Jeffrey M Vinocur <jeff at litech.org> writes:

> Great, that fixed it.

> I did have to rerun ./autogen to keep it from doing this:

>     gcc -g -O2 -I../include  -fno-strict-aliasing -pipe  
>     -I/usr/lib/perl5/5.8.6/i686-linux/CORE @PERL_WARNINGS@  -c perl.c

> and in this case it was pretty obvious that something needed to be 
> updated.  However, the fact that I had to pick this up manually means 
> there's an implicit dependency somewhere in the build process...and that 
> some subtler change could slip by without me realizing to ./autogen. 

> Is it feasible to make these dependencies explicit?  I imagine it's a bit 
> hairy to have `make` lead to the makefile being rebuilt, but perhaps just 
> have it die with a detailed message?

The problem is that there are other reasons why the timestamps might be a
little off on the various bits that go into configure (such as a fresh
checkout and a very accurate timestamp in the file system) and then you
get annoying errors.  It would be nice, but... I think it may be more
hassle than it's worth.

You can always safely run autogen after a checkout.  Usually other
projects that put the dependencies in make people run a script after
checkout to fix the timestamps, so since either way you have to run a
script, I'd rather just run autogen.  :)

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>

    Please send questions to the list rather than mailing me directly.
     <http://www.eyrie.org/~eagle/faqs/questions.html> explains why.


More information about the inn-workers mailing list