Mail Archives: geda-user/2015/12/31/14:58:11
This message is in MIME format. The first part should be readable text,
while the remaining parts are likely unreadable without MIME-aware tools.
--8323329-1982668496-1451588605=:12724
Content-Type: TEXT/PLAIN; CHARSET=ISO-8859-7; format=flowed
Content-Transfer-Encoding: 8BIT
Content-ID: <alpine DOT DEB DOT 2 DOT 11 DOT 1512312047291 DOT 15043 AT newt>
On Thu, 31 Dec 2015, Edward Hennessy (ehennes AT sbcglobal DOT net) [via geda-user AT delorie DOT com] wrote:
> I havenąt found a way to disable Xorn during make distcheck.
From an Autotools point of view, you'd want to enable all conditional
Makefiles in "make distcheck" so they are tested and distributed even if
they aren't normally enabled. It does make sense in a way that "make
distcheck" fails when a test in one of these Makefiles fails. I'm afraid
there's no easy way to fix this without fighting the tools.
That said, I completely understand that this makes testing changes to the
build system unnecessarily cumbersome. (I used to add the failing tests
to XFAIL_TESTS temporarily before running "make distcheck".) I guess the
intention behind this is that there shouldn't be failing tests in master,
at least at the time of a release.
So the options to handle this are:
(a) Don't merge the code until the defect is fixed.
(b) Have "make distcheck" fail as a reminder that the code doesn't work right.
(c) Hack the test driver so a failed test doesn't cause "make distcheck" to abort.
(d) Add the defect to the documentation as a known bug and set the corresponding tests to XFAIL.
Since this is a minor flaw, I had decided for (b), but if you think (c) or
(d) is more appropriate I wouldn't object.
> How much effort is required to fix the defects?
Basically, a new abstraction layer for "SPICE objects" would have to be
introduced in the spice-sdb backend (other backends aren't affected).
This isn't a big issue per se, but since this would make changes all over
gnet_spice_sdb.py necessary and this is the largest backend (~1250 lines),
I'd estimate 1-2 days to do it properly.
The original gnetlist presented a package as an opaque object to backends
and didn't allow them to access the individual components of which it is
composed. If a backend wanted to handle slots as individual objects (like
the spice-sdb backend does), it had to override gnetlist's "get-uref"
function so the slots were each assigned an individual refdes. In the
refactored code, this is handled differently (you can access the
"components" property of a "Package" object), so the renaming mechanism
doesn't exist, and a backend which wants to handle slots as individual
components needs to handle this itself. Unfortunately, I haven't found
the time to implement this in spice-sdb yet.
--8323329-1982668496-1451588605=:12724--
- Raw text -