Reduce Dwarf Debug Size

Mike Stump mrs@apple.com
Thu Mar 1 22:03:00 GMT 2007


On Mar 1, 2007, at 11:41 AM, Daniel Jacobowitz wrote:
> On Thu, Mar 01, 2007 at 11:24:18AM -0800, Devang Patel wrote:
>> Since then we learned many things about this approach,
>> and newest darwin tools do not use it anyway. My point was,
>> GCC lost valuable time experimenting and learning from
>> such approach in tackling not so simple issue.
>
> Apple also lost lots of time by not discussing any of this on the
> mailing lists and maintaining their own solution :-)

Your mistaken.  We've shipped many solutions in the time span that  
not much was done in mainline.  Our current solution can shave  
gigabytes off the debug information size going through the linker,  
we're happy.

> In an ideal world, the Apple tools developers would have more time  
> to work with FSF GCC (and GDB) communities so that all could benefit.

In the end, it comes down to lead, follow, or get out of the way.  I  
want gcc to lead, history will teach us if we meet that goal.  So  
far, I think we do a pretty good job, though, as times I'd like gcc  
to do better.  One thing I am certain of is, tomorrow will bring new  
demands as well as new competition.  gcc needs to spend what  
resources it has wisely and effectively to meet those demands and the  
competition.

At times, I use:

   http://www.dberlin.org/patches/patches/list

as a proxy for how effectively we're using resources, and it still  
feels to me that we could be more efficient at what we do.  I'd like  
to see us reorganize how we do what it is that we do to substantially  
reduce review time.  For example, llvm permits post checkin review.   
I do wonder if we granted post-checkin review status to people after  
they submit X good patches, if that would let us use those resources  
more effectively.



More information about the Gcc-patches mailing list