[Lldb-commits] [PATCH] D14790: cmake: add -DLLDB_ALLOW_STATIC_BINDINGS=1, defaults off

[Todd Fiala via lldb-commits]

tfiala added a comment.

In http://reviews.llvm.org/D14790#292255, @zturner wrote:

>




> What is worth arguing about, however, is that changes need to have some kind of reasonable justification.


I strongly disagree with this statement from a procedural point of view.  It is just an invalid expectation.  We are a bunch of commercial companies that work in a competitive environment.  This goes for Google, Apple, Qualcomm, Sony, etc.  I do not need a justification from each of these companies that explains items happening behind closed doors that motivate the desire for functionality in LLVM, clang, LLDB or any other open source project.

For example, Microsoft has an excellent debugger.  Truly.  World class.  Google is implementing a debugger to work on the Microsoft platform.  I don't need to know why that is the case.  It's totally fine that I don't know why.  It is not a requirement for me to know why.  Even if this requires changing support to add Python 3.5 (that doesn't break others), even if our directory structure gets 30 - 50+ characters deeper than it was before, whatever.  This is all okay procedurally speaking.  I don't have the prerogative to require knowing all those details.

I made it clear here what goal I (and by I, we're talking Apple now) am trying to achieve:
http://lists.llvm.org/pipermail/lldb-dev/2015-November/008802.html

In particular:
"The primary goal here is to remove the requirement of having swig on the
system (e.g. for builders and most developers)"

You keep asking the question and not liking the answer, but the goal is to "not require swig."  The approach has been watered down so that this only happens when opted in, but that is the goal.  And this is Apple's goal.  You may not like the goal, and I am not going to try to persuade you to like it.  But that is the goal.

At this point, it is entirely opt-in, and off by default.  Nobody other than Apple has to maintain the static bindings.  So it is our goal, we implemented a way to do it, and we are responsible for maintaining it.  If a by-product of this breaks somebody who didn't opt into this, then I think that is something I'd need to look at us addressing.

But no, asking for repeated justification for *why* that is Apple's goal beyond what I've already said (i.e. making it simpler for a casual developer to get a machine that can build lldb) seems both an unreasonable expectation and completely unnecessary to discuss.  Again, I'm happy to rip out the manner in which we're achieving that if something considered better comes along (which your bindings-as-a-service idea sounds like a potentially promising avenue to look at).  But I don't see there being a requirement to argue a goal that doesn't hamper the existing developer community.


http://reviews.llvm.org/D14790