On Tuesday, January 5, 2016, Drew Crawford via swift-build-dev < swift-build-dev@swift.org> wrote:
I think I understand better what you are trying to accomplish. It is not
a bad way to go about it, and I think my comments were not as constructive
as they should have been.
I sent a PR that may clarify your motivation so as to head off objections
like mine. I think my reaction to the document was largely based around a
misunderstanding about the motivation. The title is "Package Manager C
Language Target Support", and so that *sounds* like it is a proposal
about C language support as a whole, which immediately raises questions
about whether Foundation et al can be someday made to fit in its box (e.g.
via extending it etc.) Whereas if we are clearer that this is *merely
one path to C language support, and this is the first one on the whiteboard* then
I think it is a lot stronger. It does solve a problem, I have a better
idea of what that problem is, and the fact that it doesn't map well onto my
problems is no longer troubling.
I think your language is more polarizing than is due here, and I would
encourage focusing on a technical argument rather than a judgement like
"the entire value is debatable".
I apologize for that.
For background, when the topic of native C support vs external build
systems has come up in the past, there's been an implicit value judgment
that building C with swiftPM is the "preferred" way, and using an external
build system is the "compromise" way, and when we set up that value
hierarchy it is necessarily the case that the two ideas are in conflict.
I'm coming around to the idea that actually these are totally unrelated
problems, they only looked similar at first glance, that neither is any
better or worse than the other, that we need to support both, and so we
should just give both of them permission to be excellent in their own way,
and that we should not try to shoehorn one of them into the problems that
the other one is better at.
Your proposal is clearly going to be better at the "few files of C code"
problem, it was a bad idea for me to overcomplicate it with the problems of
established C projects when they can be better served by handling them with
an entirely separate solution.
On Jan 4, 2016, at 6:48 PM, Daniel Dunbar <daniel_dunbar@apple.com > <javascript:_e(%7B%7D,'cvml','daniel_dunbar@apple.com');>> wrote:
Hi Drew,
First off, I believe that the right way to handle something like libsodium
is to have support for building targets using an external build system. I
see that as a separate and independent feature from this proposal though.
If you are interested in working on that feature I would love to discuss it
more on a separate thread (I already mentioned it explicitly on another
thread I am unable to find right now) -- it is a feature I would really
like to see us have but don't have the bandwidth to tackle at the moment.
This proposal is very specifically targeting the desire to be able to
write and build new C code as part of Swift packages, it is not designed to
support importing large existing projects. While I do hope that it will
feature creep over time to allow more and more C projects to fit within the
supported conventions, I also expect that to be a long incremental process.
At this stage of the project, I would encourage looking at new proposals
and features from a perspective of "does this add a useful new capability"
and "is this in line with our goals" rather than "does this solve my
immediate need X".
More comments in line...
On Jan 3, 2016, at 12:17 AM, Drew Crawford via swift-build-dev < > swift-build-dev@swift.org > <javascript:_e(%7B%7D,'cvml','swift-build-dev@swift.org');>> wrote:
You keep referring to "'real' projects" as a proxy for the individual
project you want support for; while there's a lot to be said for real-world
use cases, I don't think this proposal's direction should be dictated by
just libsodium.
Nothing about this is reductive or specific to libsodium.
* libdispatch has a build manual
<https://github.com/apple/swift-corelibs-libdispatch/blob/master/INSTALL> that
runs over 100 lines and involves checking out *7* other repos.
* Foundation (incl CoreFoundation) has a python-based build
<https://github.com/apple/swift-corelibs-foundation/tree/master/lib> system
that runs to 1900 lines
* libjpeg is "the" example of a C dependency in our documentation, and its
build system includes such goodies as choosing a memory manager
<CVS Info for project libjpeg; or
configuring libpng.
And these are just the projects that *we* are associated with!
These are projects I would expect to be tackled with an "external build
system" feature, not this proposal.
There are C projects that would benefit from modularization, header
auditing, and cleanups that Swift and swiftpm would bring to it.
Swift and swiftpm don't do anything of the kind. C developers do. All we
can do is try to impose new requirements on C developers. And who is
volunteering to implement those requirements?
It seems to me that if our new requirements are so amazing, it should be
easy to convince a few projects to sign on to repackage. libdispatch and
Foundation are *our* projects; the bar is so low we're practically
cheating. Are they going to switch to this as their build system? I don't
know who makes this decision, but it seems like an important question to
ask.
While I think that "support libdispatch and Foundation" are good long term
goals and useful reference points for what features are still needed to get
there, it isn't the immediate goal here.
No Swift project could be built with swiftpm when it was introduced
without repackaging. I don't see why C support should be held to a
different standard.
Because we're designing a package manager for the Swift language, not the
C language. C has had build systems for decades. We're not going to just
waltz in with a new standard for a 44-year-old language and everybody
switches the next day. This is https://xkcd.com/927/\.
We can all dream, right? :)
On Jan 2, 2016, at 8:11 PM, Zach Waldowski via swift-build-dev < > swift-build-dev@swift.org > <javascript:_e(%7B%7D,'cvml','swift-build-dev@swift.org');>> wrote:
*I think an important feature of any C target proposal is that there will
actually exist C targets which can be built under the proposal*. Until
there are C people coming out of the woodwork saying "sure, I will
repackage my software this way" I think the entire value is debatable.
I almost couldn't disagree more. No Swift project could be built with
swiftpm when it was introduced without repackaging. I don't see why C
support should be held to a different standard.
I agree with Zach here. The purpose of the proposal is to add support for
new code written in C designed to integrate with other Swift code. The
merits of the conventions should be weighed about how easy it is to write
that code, and what the implications for maintenance (on both the package
manager and the code itself) are.
And I do not see realistically how we are ever going to support a project
like libsodium, except calling out to automake.
A potential solution (one of many possible) would look a lot like how
people generate Xcode projects for C build systems today; hand-tuning
config.h headers and such. I know many people who will go to ungodly
lengths to avoid the inevitable nightmare automake causes in a
source-distributed dependency.
IMO something like that is a much, much better direction in the
short-term, and once we have done the first step of "packaging" those
software via automake we will have "real" C projects in our package manager
and we can design our C support around the concerns of real projects
instead of imaginary ones.
There are C projects that would benefit from modularization, header
auditing, and cleanups that Swift and swiftpm would bring to it. C projects
are massively disorganized because build systems are a ridiculous
hodgepodge; we didn't be subject to that long tail of good and bad
decisions.
I don't think automake support would be a silver bullet at all, and
contradict with many goals of swiftpm and llbuild to boot. Even targeting a
really small subset of automake projects what liberties would unnecessarily
complicate the project, and then there'd be the projects it doesn't
support. (Oh? Wait? What version of the tools? Oh, from trunk? Oh, does
the project take any liberties with its own organization? God help us when
we start talking about C++…)
"imaginary" is a reductive way of phrasing the problem space. You keep
referring to "'real' projects" as a proxy for the individual project you
want support for; while there's a lot to be said for real-world use cases,
I don't think this proposal's direction should be dictated by just
libsodium.
Zachary
On Sat, Jan 2, 2016, at 04:57 PM, Drew Crawford via swift-build-dev wrote:
Thanks for directing me to this, I missed it.
Most projects will not conform to these conventions.
Giggle. Kind of an understatement, no?
Like, okay. Here is a project I'd like to package. (Read: I do package
it, with features not in mainline swiftPM.)
GitHub - jedisct1/libsodium: A modern, portable, easy to use crypto library.
Let's take a look at how this package realistically builds:
Here is how I would expect these things to be tackled:
* It has tests ("make check")
I don't have any particular plan for this one.
* It has various --enable-foo flags
I expect that eventually the package manager will grow some level of
support for this kind of thing (similar to Cargo's "features" --
Page Moved).
* It swaps in special implementations depending on if you have AMD64
<https://github.com/jedisct1/libsodium/blob/master/src/libsodium/Makefile.am#L162>
or AVX instructions
<https://github.com/jedisct1/libsodium/blob/master/src/libsodium/Makefile.am#L145>
or SSE2
<https://github.com/jedisct1/libsodium/blob/master/src/libsodium/Makefile.am#L229>
etc.
* The optimization level is tuned on a per-architecture basis
<https://github.com/jedisct1/libsodium/blob/master/dist-build/android-armv7-a.sh#L3>
I expect that both of these would become possible with small enhancements
to allow customization via the manifest file. Again, this is just the
initial feature.
* They build (also) on Windows. They're not changing how they're packaged
for "SwiftPM, the Mac/Linux build system".
I anticipate that we will eventually support alternative source layouts
via customization in the manifest file (mostly to change where the sources
are found, and the way headers are located and treated).
* Oh and this is cryptography code. Do you *really* want to touch it?
It is a non-goal of this proposal to support projects which "I don't want
to touch".
*I think an important feature of any C target proposal is that there will
actually exist C targets which can be built under the proposal*. Until
there are C people coming out of the woodwork saying "sure, I will
repackage my software this way" I think the entire value is debatable.
I think your language is more polarizing than is due here, and I would
encourage focusing on a technical argument rather than a judgement like
"the entire value is debatable". This proposal will clearly allow packages
to add small bits of C code which are used by other targets. Unless you
believe that is incorrect (and if so, please present a technical argument
for it), then to me that is a valuable capability (and if you disagree,
then please present a technical argument for it).
- Daniel
Getting signoff from libdispatch/CoreFoundation is necessary but not
sufficient to clear that hurdle. I would think getting the other C deps in
our own project family to repackage would be "table stakes" for any new C
build system. The real test are projects that are third-party and less
friendly.
And I do not see realistically how we are ever going to support a project
like libsodium, except calling out to automake. An automake solution
coincidentally supports both libdispatch and CoreFoundation right now. IMO
something like that is a much, much better direction in the short-term, and
once we have done the first step of "packaging" those software via automake
we will have "real" C projects in our package manager and we can design our
C support around the concerns of real projects instead of imaginary ones.
On Jan 2, 2016, at 11:00 AM, Daniel Dunbar via swift-build-dev < > swift-build-dev@swift.org > <javascript:_e(%7B%7D,'cvml','swift-build-dev@swift.org');>> wrote:
Happy 2016!
I am working on an initial proposal for adding support for C language
targets to the Swift package manager, and am interested in feedback:
https://github.com/ddunbar/swift-evolution/blob/master/proposals/NNNN-swiftpm-c-language-targets.md
Some TL;DR:
- The proposal defines a basic convention for pure C language targets (no
Swift/C mix and match, but other Swift targets can use the C targets).
- This is just intended to be the minimal initial feature, there will be a
lot of add on work which I expect should be tackled in follow on
PRs/proposals.
- The proposal doesn't try and outline every single nitty detail (e.g.,
exactly what C++ standard we will compile with). My intention is to pick a
sensible default at implementation time and refine incrementally.
Unless there are serious objections, I am hoping to hope to land this
proposal soon and start work on the feature shortly after.
Cheers,
- Daniel
_______________________________________________
swift-build-dev mailing list
swift-build-dev@swift.org
<javascript:_e(%7B%7D,'cvml','swift-build-dev@swift.org');>
https://lists.swift.org/mailman/listinfo/swift-build-dev
*_______________________________________________*
swift-build-dev mailing list
swift-build-dev@swift.org
<javascript:_e(%7B%7D,'cvml','swift-build-dev@swift.org');>
https://lists.swift.org/mailman/listinfo/swift-build-dev
_______________________________________________
swift-build-dev mailing list
swift-build-dev@swift.org
<javascript:_e(%7B%7D,'cvml','swift-build-dev@swift.org');>
https://lists.swift.org/mailman/listinfo/swift-build-dev
_______________________________________________
swift-build-dev mailing list
swift-build-dev@swift.org
<javascript:_e(%7B%7D,'cvml','swift-build-dev@swift.org');>
https://lists.swift.org/mailman/listinfo/swift-build-dev