[lldb-dev] [RFC] Delete the "testcases" symlink

[Zachary Turner via lldb-dev]

It's definitely possible to (at worst) shorten the path if we say we don't
care about supporting other languages (i'm fine saying that for now unless
someone comes along with a serious committment to support and maintaining
other bindings).  So we could drop packages/Python for example.

I also remember the whole third party debacle now.  That was (IIRC) the
reason for the packages indirection, because we want both LLDB scripts
(i.e. stuff in the scripts/ folder) as well as test cases and test
infrastructure to all be able to share the same code.

Things get really nasty when you start putting stuff in different trees,
Python isn't really designed to work well in this scenario, it expects all
code to share a common root, which is why it was difficult to have
testcases and package code in a different tree.

If someone finds a way to make it work though, more power to them :)



On Thu, Dec 15, 2016 at 10:58 AM Jim Ingham <jingham at apple.com> wrote:

> Yes, writing that long path every time you run a test case was indeed the
> reason for that symlink.  I stood it for a couple of weeks before it drove
> me round the bend!  You want it to be a real path and not just have the
> test suite accept partial paths because it is really handy to use the
> shell’s filename completion.
>
> Anyway, if there’s some compelling reason to remove the link, I’ll just
> remake it by hand in all my checkouts, not too much of a hardship.  But it
> would be much better if we could remove that long path and move the test
> cases back into the test directory.
>
> Jim
>
> On Dec 15, 2016, at 9:45 AM, Zachary Turner via lldb-dev <
> lldb-dev at lists.llvm.org> wrote:
>
> The reason they were put into the packages directory in the first place is
> because this forms a proper python package so that more code can be shared
> amongst test cases.  I don't know why the symlink was ever created in the
> first place, sounds like that was probably put in place by someone who
> thought packages/python/lldbsuite/test was too much to type (hard to argue
> with, tbh).  In any case, as long as you can maintain the package structure
> so that a testcase can write "from support import fs" then you can
> re-arrange stuff however you like.
>
> I also think some people had expressed interest in writing bindings for
> other languages, but I haven't heard anything about that in some time, so
> until I see more effort on that front a concrete proposal, we don't need to
> worry about it.
>
> On Thu, Dec 15, 2016 at 9:38 AM Chris Bieneman via lldb-dev <
> lldb-dev at lists.llvm.org> wrote:
>
> So, I want to delete the testcases symlink from the LLDB tree, and move
> the tests out of the packages directory into tests/testcases.
>
> A few reasons why I want to do this:
> (1) I can't imagine symlinks in SVN is something that works well on
> platforms that don't support symlinks (like Windows)
> (2) It trips up scripts and tools that try to operate on the files without
> calling realpath (while I know I could fix those, it is easier to just fix
> this)
> (3) It seems to me that it is completely unnecessary. In lldbsuite's
> __init__.py we already find the test cases by path from the lldb checkout
> root, so having them live in the python packages just seems odd.
>
> Thoughts?
>
> Is there a reason why they need to live in the packages directory that I'm
> unaware of?
>
> -Chris
> _______________________________________________
> lldb-dev mailing list
> lldb-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-dev
>
> _______________________________________________
> lldb-dev mailing list
> lldb-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-dev
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/lldb-dev/attachments/20161215/2b309568/attachment-0001.html>