commit 875406c9d1211fd88e3e7e1ee5349cba41a7becf
Author: Peter TB Brett <email address hidden>
Date: Wed Sep 25 22:17:56 2013 +0100
Update NEWS and version information for 1.8.2.
commit 1ae405d39a3822875d3dbc5f20f8152bf172bed7
Author: Peter TB Brett <email address hidden>
Date: Wed Sep 25 22:24:51 2013 +0100
Update translations.
commit 35e53ee0b30b5395a6007eb235c6479656be2d26
Author: Peter TB Brett <email address hidden>
Date: Wed Sep 25 22:32:57 2013 +0100
Update README with new Automake requirement.
(imported from commit 82c5b984ca51720ea8078aed326c31b37e4e3e33)
commit 1dd3f5d77dc013624f6848801f4885c485595425
Author: Peter TB Brett <email address hidden>
Date: Wed Sep 25 22:34:27 2013 +0100
build-sys: Don't use Automake test driver unless actually necessary.
Most of the tests in geda-gaf actually provide their own test drivers
in the form of a shell script. It's actually better just to run them
directly in a check-local rule than to use the Automake test driver
(which eats all the output now that parallel tests are being used).
(imported from commit b7a373f8846fa91c92e4a628e0691b7f865af3e0)
commit 9013d967a865f35c0f946e30c0f41073cc606e5f
Author: Peter TB Brett <email address hidden>
Date: Wed Sep 25 22:30:23 2013 +0100
build-sys: Use parallel tests for Scheme API.
(This makes the testsuite work again).
(imported from commit 0f3c4785ec997b9c04bdfc695192811f6c5eba43)
commit 30811530b71cb2079d812d7e2a8e444fb745d611
Author: Peter TB Brett <email address hidden>
Date: Wed Sep 25 22:29:20 2013 +0100
build-sys: Use parallel tests with coloured output.
Note that changing this option means that `make check' and `make
distcheck' will fail until all the tests have been ported over.
(imported from commit c518fc9dc04f25c9cfbfed5d32444eb3168ed506)
commit 64d99b7efd1494c2a441061cb2bbc4bbc57d6a46
Author: Peter TB Brett <email address hidden>
Date: Wed Sep 25 22:28:42 2013 +0100
build-sys: Use GUILE_PKG_NAME instead of GUILE_PKG.
We previously stored the name of Guile's pkg-config module in a
variable called GUILE_PKG. In recent releases of Guile 2.x, a
GUILE_PKG autoconf macro has been added which conflicts with this. So
we need to use a different variable name instead.
(imported from commit 1c2f215ffa639f543a0ca1997c17cfbeffb76793)
commit 4cdb25cc125e8390ae2dc0d40cae2839372e848a
Author: Peter TB Brett <email address hidden>
Date: Wed Sep 25 22:28:31 2013 +0100
build-sys: Ignore new build-tools installed by Automake 1.13.
(imported from commit 4d5db45a89ad35b37fcb894b27801f953ce28e94)
commit 99be36dcd627b0326eb9e0f86834c6979fa88227
Author: Peter TB Brett <email address hidden>
Date: Wed Sep 25 22:28:13 2013 +0100
build-sys: Require Automake 1.11 for developers.
A newer version of Automake is required in order to ensure that the
parallel test driver is available.
In newer versions of Automake, this new driver is used by default and
it's necessary to use
AM_INIT_AUTOMAKE([serial-tests])
in order to get the (deprecated!) serial test driver currently used
throughout the geda-gaf package. This doesn't provide a solution: at
the moment, we have the problem that 1) newer versions of Automake
choke on the test suite without the "serial-tests" option; 2) older
versions of Automake choke on the "serial-tests" option.
The only option available to retain the ability for all
currently-active developers to run the test suite is to port it to the
parallel test driver. 1.6.0 is too old; 1.11.0 was released in 2009
and seems to provide a reasonable balance between maturity, widespread
availability and features. Note that users without Automake >= 1.11.0
will still be able to build tarballs, but will need to upgrade to
build from git.
(imported from commit 0dd51a48d23c2f45ce1040fceadf653dbba008e7)
commit f5cda460b09a7af60d4a00a469321c524816af3e
Author: Peter TB Brett <email address hidden>
Date: Wed Sep 25 22:40:12 2013 +0100
gnetlist: Don't use gettext macros.
Commit dcd9b1c6eb47 introduced a translatable string to gnetlist, but
in the stable-1.8 branch gnetlist is not yet internationalised.
commit dcd9b1c6eb472b80117d945896c56092acfdeac7
Author: Peter TB Brett <email address hidden>
Date: Sun Sep 22 10:12:21 2013 +0100
gnetlist: Error on clashes with netattrib-generated refdes.
Exit with an error if a refdes is found in input files that conflicts
with autogenerated netattrib-related refdes.
Closes-bug: lp-1226246
commit c58247b598517b875d8cb2b80eb812e92ee9c87f
Author: Peter TB Brett <email address hidden>
Date: Sun Sep 22 10:11:23 2013 +0100
gnetlist: Safer use of nettattrib-generated refdes.
Two-pronged attack to make gnetlist's use of autogenerated "POWER"
refdes less likely to cause a problem.
1) Use a longer, more specific prefix string. This will make it much
much less likely that a collision with a user net name will occur.
2) When testing for autogenerated refdes, test for the exact prefix
using strncmp() instead of just checking that it appears somewhere
in the string.
Affects-bug: lp-1226246
commit ed432e7e2ff1c38d2b04608b2d0287e137788c1e
Author: Peter TB Brett <email address hidden>
Date: Sat Nov 24 22:06:09 2012 +0000
libgeda: Refuse to connect nets inside symbols with other nets.
A libgeda crash bug was caused by the pathological use case of putting
nets inside symbols. Since the file format does not specifically
forbid this, libgeda must cope with this gracefully. Specifically,
o_net_consolidate() was causing a crash when attempting to merge a net
inside a symbol and a normal net. The underlying cause was found to
be that the connection tracking system did not cope properly with
encountering connectable objects other than pins inside symbols.
Merely fixing the net consolidation algorithm exposed other bugs (for
example, nets inside symbols should NOT rubberband!)
This patch adds logic to s_conn_update_line_object() as follows:
1) If two objects are both not inside symbols, they may connect.
2) If two objects are inside the same complex, they may connect.
3) If an object is inside a complex, it must be a pin to connect to
objects outside the complex.
Closes-bug: lp-994361
(imported from commit c18a0826fcd43dc8471f98e3a092d3144fcb6918)
