[Top][All Lists]

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

Re: Storing generated files (was: Re: Make cvs ignore some changes in fi

From: Noel Yap
Subject: Re: Storing generated files (was: Re: Make cvs ignore some changes in files)
Date: Thu, 22 Aug 2002 12:49:59 -0700 (PDT)

--- Mike Ayers <address@hidden> wrote:
>       1)  The generation process itself is too expensive.
>  Hardware (FPGA 
> and ASIC) design tools and 3D graphics come to mind,
> where generation 
> can require several days on a cluster.  One cannot
> simply expect each 
> developer to generate their own result file, unless
> one owns 
> controlling interest in the server supplier.

I agree.

>       2)  The generation will not necessarily produce the
> same result.  A 
> good example is lex/flex and yacc/bison generation. 
> Lex and yacc are 
> different tools on most of the machines on which
> they occur, yet tend 
> to produce better (and, as I recall, more
> portable(?!)) results than 
> flex/bison on those systems.  A project I worked on
> would generate 
> files using lex/yacc on a specific platform and
> archive them into the 
> sources.  Those building the project would compile,
> and most would get 
> satisfactory results.  Those who did not could clean
> out the generated 
> files and rebuild; this usually worked.  We could
> not require 
> generation of these files, as not all systems had
> lex/yacc or flex/bison.

IMHO, this isn't good CM.  If you want developers to
use the same generated file, they should be using the
same generation tool.  If this isn't possible, it
becomes case 3) below.

>       3)  The generation tool is not available on all
> supported platforms.

I agree.

>       4)  The generation tool is nondeterministic.  There
> are two cases of 
> this that I know of.  First, there is Altera's FPGA
> compiler (at least 
> last time I checked).  Next is cryptography:  one
> would want to start 
> by generating random seeds.  However, for test runs,
> one would like to 
> archive a seed file so that results are consistent.

I'm pretty sure there's nothing non-deterministic in
the computer world (except maybe for some cryto
hardware) so I'll assume you couldn't think of a
better word (nor can I).

Anyway, I can't comment on the compiler since I'm not
familiar with it.

As for a random seed for testing, I'd say this is the
same as using vi (and manual typing) or recording a
feed to generate the data file.  If it's used for
testing, the file shouldn't be changing that often.


Do You Yahoo!?
HotJobs - Search Thousands of New Jobs

reply via email to

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