[Omake] SVN Commit: OMake Build System (Rev. 10740)

[Jason Hickey]

On May 14, 2007, at 10:36 AM, Aleksey Nogin wrote:

> On 14.05.2007 10:05, Aleksey Nogin wrote:
>
>> A temporary hack of 0.9.8.2 only: in AST-TO-IR conversion,  
>> recognize the
>> ".STATIC:" section header as if it was "static. =". This is a HACK  
>> that is
>> meant to add a limited form of forward-compatibility with the  
>> upcoming
>> 0.9.8.3+ syntax.
>> ----
>> Changes:
>>     +14 -1	omake-branches/0.9.8.2/src/env/omake_ir_ast.ml
>>  A hyperlinked version of this commit is available at
>> http://svn.metaprl.org/commitlogs/omake/ 
>> 2007-05.html#07/05/14.10:05:28
>
> Jason,
>
> Is this a reasonable thing to do? It's obviously a big HACK, but  
> not a one that we'd have to tolerate for long. I think if the  
> syntax is about to change even within the 0.9.8.x version series,  
> we'd better add at least some minimal forward compatibility...

I think its reasonable.

>
> On the other hand, I am not sure I completely agree with the syntax  
> you chose to implement. The current syntax is
>
> .STATIC: dependencies
>    definitions (implicitly declare?)
>
> while the one we've discussed before (and I still think might be a  
> better idea) is
>
> .STATIC: declarations: dependencies
>    definitions

I understand, and I've waffled on this one.

The standard syntax for rules is one of the following, except  
for .SCANNER rules (which hack the order).

     <targets>: <patterns>: <dependencies>
     <targets>: <dependencies>

We could enforce the 3-place rule, but we need an easy syntax for the  
case where all the variables are exported and there are no  
dependencies (the common case).  The export should have the same  
syntax as <exports> in "export <exports>".  But the following is IMHO  
very ugly for the common case.

     .STATIC: :

I believe that the second most common case is to have dependencies,  
and still export everything.  Again, I don't like the syntax...

     .STATIC: : foo.txt
         TXT = $(cat foo.txt)

In the end, the only thing I am sure about is that I don't like  
the : : sequence.

Jason

--
Jason Hickey                  http://www.cs.caltech.edu/~jyh
Caltech Computer Science      Tel: 626-395-6568 FAX: 626-792-4257