[LLVMdev] "ninja update" not working.

[Chandler Carruth]

On Mon, Jun 25, 2012 at 3:21 AM, Csaba Raduly <rcsaba at gmail.com> wrote:

> On Mon, Jun 18, 2012 at 11:46 AM, Chandler Carruth  wrote:
> > On Mon, Jun 18, 2012 at 2:32 AM, Mikael Lyngvig  wrote:
> >>
> >> Hi,
> >>
> >> The docs state that you can do a "make update" to update all top-level
> and
> >> embedded projects from Subversion.
> >
> >
> > This is a feature of the configure+make build system, and not a feature
> of
> > the CMake build system. I also don't have any real desire to support it
> as
> > part of the CMake build system, I think it was always a flawed idea to
> > implement version control commands as part of a build system.
> >
> > For example, I never, *ever* to in-tree builds, and a usually have many
> > different builds associated with a single source repository. This makes
> the
> > semantics of 'make update' quite confusing.
>
> I can't see why it's confusing. I have only one source tree and any
> number of build trees; "make update" in each build tree updates the
> source tree.


No other part of building mutates the source tree. Thus, for all other
things I do in the build tree, I can do them in N different built trees
concurrently without issue. Deviating from this is a dramatic and confusing
deviation.

What happens when the source tree is on a read-only (network mounted)
filesystem?

What happens when somehow someone accidentally typos a dependency from some
build rule to the 'update' build rule?


I don't see any really good way to integrate manipulation of the version
control system from within the build system, nor do I see really compelling
advantages to doing so, or really compelling problems with alternate
solutions.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20120625/436f0f98/attachment.html>