Wireshark-dev: Re: [Wireshark-dev] [Wireshark-commits] rev 38304: /trunk/epan/dissectors/ /trun
From: Jeff Morriss <jeff.morriss.ws@xxxxxxxxx>
Date: Fri, 07 Oct 2011 12:42:03 -0400
Balint Reczey wrote:
On 10/07/2011 04:26 PM, Jeff Morriss wrote:
Balint Reczey wrote:
On 10/07/2011 04:07 PM, Jeff Morriss wrote:
Gerald Combs wrote:
On 10/4/11 1:11 PM, Jeff Morriss wrote:
Balint Reczey wrote:
Let's make Python a requirement now.
I fully support the idea of not keeping generated files in SVN and in
the release tarball.
I know it's been a while, but...  Any other thoughts/votes on this
topic?

It's fine with me.

One downside with this is that it takes (at least on my not-that-old
home PC) about 3.5 minutes to build all the ASN.1 dissectors ("time make
-C asn1").  Since the ASN.1 dissectors are all in subdirectories (and
each subdir has only one target) the build does not parallelize ("make
-j X") well.

It probably could be made more parallel through some ugliness (making
what is currently a recursive build non-recursive).

Thoughts?
We can make it parallel without converting them non-recursive.
I'm in for improving the parallelism.

How?  (At the moment I'm looking only at automake, but I suppose
eventually Windows and CMake must be considered.)
I was thinking about using build-stamps and inter-directory dependencies without implementing a full non-recursive make system, but I have to investigate the current situation to be sure.

In the meantime maybe we could simply speed up the asn2wrs script, but I haven't done any profiling on it yet.

Right now I'm getting errors even with -j1. :-(

I have it working except for asn1/x721 which, well, doesn't actually have a dissector in it yet (need to find a way to make it do nothing for "make all"). I can check in my changes--thus far just cleanup and putting the asn1 dissectors straight into epan/dissectors/ tonight or this weekend. (I think the changes I made will port over easily to Windows/nmake.)