Kill a hack for explicit isinstance(). We've got just two methods,
so doing full getmro doesn't seem to make too much sense.
Anyway this changes semantics slightly, so now we rely on exact
inheritance rather than name (I don't have good answer for those)
--HG--
branch : trunk
that fijal didn't move py.test.Item|Function to the py.test.collect namespace
at the time. a pypy fix will follow shortly after i merged the dist.
--HG--
branch : trunk
setup()
try:
...
finally:
teardown()
and not to do the setup() within the try,
because if that raises, teardown() will
likely go wrong as well etc.
--HG--
branch : trunk
from a default rsync_root (checking
for rsync_root == topdir is not safe
as the topdir may be explicitely contained
in the rsync_root list)
--HG--
branch : trunk
specified, the config.topdir is transfered but it is
transferred to the "remotepath.join(topdir.basename)"
(not actual code) to avoid random such rsyncs to
destroy/affect remote filesystem state.
--HG--
branch : trunk
do not get it.
Right now there is no way to override default session other than
from py.__.test import config; config.TerminalSession = 'name_to_import'
which is scary. Holger please take a look.
--HG--
branch : trunk
in the FileBox affects --dist testing in ways i
can't fully understand at the moment (i would
think it only affects --boxed test runs). So
we are back to share tempdirs when running boxed
(causing failures elswhere). We care later, i guess.
--HG--
branch : trunk
because getpid() is different per thread
(and if e.g. run via execnet in threads ...
it creates new tempdirectories all the time)
let's consider this sometime else
--HG--
branch : trunk
want to rsync to home dir usually)
* Make a flag rsync_flag in hostinfo which tells whether to rsync
or no, semantics are that in case of localhost we do not rsync
and in case of localhost: we do.
--HG--
branch : trunk
"config.option.boxed" indicates now boxed tests
and RSession.fixoptions takes care to honour
dist_boxed accordingly. So you can do
if not py.test.config.boxed:
py.test.skip(...)
i also fixed the documentation.
--HG--
branch : trunk
for that. Basically - callbacks with lexical scope variables are usually
(depends of coz) called with variables after the for loop has finished.
--HG--
branch : trunk
something is odd regarding the messages, i think
(read the diff, it's a simple change, i guess
we actually want to generate an Event also for
the web reporter at some point, i think,
but we could live with this for the merge IMO)
--HG--
branch : trunk
with the real world (through writing or creating
pytestcache-XXX files), semantic change: upon
gateway initialization the remote path is not
automatically created (this will be done later
by rsync anyway)
--HG--
branch : trunk
have the other side chdir()ed to home,
thus generalizing 37971 with respect to platform
support and making it independent from gateway
implementation details.
--HG--
branch : trunk
code for re-scheduling (i guess)
fijal: in the diff you'll find a XXX fijal,
i changed the meaning of a test, wasn't sure
about it. can you check?
--HG--
branch : trunk
provided in the exportdefs, copied to the namespaces (from whereever they
come), and added docstrings to all exposed namespaces (except for _thread for
now).
--HG--
branch : trunk
intended at this point:
* HostManager and HostRSync are now acting
more locally, also easier to test.
* HostInfo deals with setting up gateways now
* HostManager, HostRSync and HostInfo are
all tested now in test_hostmanage.py
(and do not involve a full startup of RSessions)
* for rsyncing, the original directory structure
(relative to config.topdir) is preserved on the
other side, this makes "dist_rsync_roots" relatively
clean now (but it doesn't pick up things on the fly,
only initialises at the beginning)
* added lots of tests
* removed more occurences of pkgdir
* streamlined and simplified some tests
* removed lots of tests that do not appear to test
specifically enough (and caused trouble for
the refactoring)
* removed lots of (but not all, i guess) test-specific
functionality in hostmanage.py and a bit in rsession.py
* removed HostOptions() in favour of rather directly
accessing config values
--HG--
branch : trunk
* unifying IO capturing methods
* py.io.StdCapture and py.io.StdCaptureFD
(and both have a classmethod 'call' that is
a shortcut for capturing output while
executing a function)
* removing lots of duplicate code
* providing some examples in py/doc/io.txt
at least tests on win32 and linux seem
to pass all for me.
--HG--
branch : trunk
to the original (patched) file descriptor. Also made that the capturing object
is passed to apigen.py's build() function (from rsession.py), which uses the
new methods to print progress information.
--HG--
branch : trunk
should kill unnecessary frames. Right now tracing is done only for
py.test.Function objects, I don't know what to do with sth else, let's
leave it alone.
--HG--
branch : trunk
* added some tests for implied and conflicting options
* make all Session objects inherit from the base session
* small cleanups with respect to test setup
* separate tests a bit
--HG--
branch : trunk
Cleaned up the tests a bit and re-enabled some (sucky tests are better than
no tests at all, I guess... :| ), re-enabled a debug print after complaints
from cfbolz ;) but this time it prints to stderr (to avoid ReST pollution).
--HG--
branch : trunk
of most of the pkgdir and changed to use get_collector_trail.
Some tests need to be rewritten or killed, right now they're skips.
There are no checks (yet) whether this works with topdir being something
else than previous pkgdir, we'll see...
--HG--
branch : trunk
revisions 36936 to 37267:
------------------------------------------------------------------------
r36986 | fijal | 2007-01-19 13:05:58 +0100 (Fri, 19 Jan 2007) | 4 lines
Improve the behaviour off cleanup and channel closing, at least on slave side.
Some cleanups, got rid of some hacks (in favor of other hacks, but posix
is a bit hackish anyway)
------------------------------------------------------------------------
r36943 | fijal | 2007-01-18 16:09:30 +0100 (Thu, 18 Jan 2007) | 3 lines
at least LSession should work now. Did not touched yet the remote options stuff,
but some simplifications were done.
------------------------------------------------------------------------
r36937 | fijal | 2007-01-18 14:22:54 +0100 (Thu, 18 Jan 2007) | 2 lines
Create a branch for further config cleanups.
------------------------------------------------------------------------
--HG--
branch : trunk
revisions 36936 to 37267:
------------------------------------------------------------------------
r37262 | hpk | 2007-01-24 14:55:00 +0100 (Wed, 24 Jan 2007) | 2 lines
fix failing test
------------------------------------------------------------------------
r37015 | fijal | 2007-01-19 17:24:32 +0100 (Fri, 19 Jan 2007) | 2 lines
Major cleanup of hostmanage. Got rid of global functions in favor of some better-structured classes.
------------------------------------------------------------------------
r36974 | fijal | 2007-01-18 22:32:04 +0100 (Thu, 18 Jan 2007) | 2 lines
Fix the test to actually perform RSync of pylib itself and a test if that was true.
------------------------------------------------------------------------
r36970 | hpk | 2007-01-18 20:33:16 +0100 (Thu, 18 Jan 2007) | 6 lines
some little cleanups - getting rid of we_are_remote
which is not used anywhere as far as i can see.
maciej: please check that i didn't mess anything
up, but for me the same tests as before pass
------------------------------------------------------------------------
r36968 | hpk | 2007-01-18 20:00:21 +0100 (Thu, 18 Jan 2007) | 2 lines
fixing some imports
------------------------------------------------------------------------
r36966 | fijal | 2007-01-18 19:36:02 +0100 (Thu, 18 Jan 2007) | 2 lines
Add try: finally:
------------------------------------------------------------------------
r36965 | fijal | 2007-01-18 19:35:00 +0100 (Thu, 18 Jan 2007) | 2 lines
Fix the test.
------------------------------------------------------------------------
r36963 | fijal | 2007-01-18 19:21:43 +0100 (Thu, 18 Jan 2007) | 2 lines
Fix remaining tests and make strange hack with parse/reparse.
------------------------------------------------------------------------
r36962 | fijal | 2007-01-18 19:06:13 +0100 (Thu, 18 Jan 2007) | 3 lines
Huge refactoring. Got rid of remote_options and session_options, two tests
still fail, so it's intermediate checkin.
------------------------------------------------------------------------
r36943 | fijal | 2007-01-18 16:09:30 +0100 (Thu, 18 Jan 2007) | 3 lines
at least LSession should work now. Did not touched yet the remote options stuff,
but some simplifications were done.
------------------------------------------------------------------------
r36937 | fijal | 2007-01-18 14:22:54 +0100 (Thu, 18 Jan 2007) | 2 lines
Create a branch for further config cleanups.
------------------------------------------------------------------------
--HG--
branch : trunk
revisions 36936 to 37267:
------------------------------------------------------------------------
r37065 | arigo | 2007-01-20 17:15:10 +0100 (Sat, 20 Jan 2007) | 3 lines
Choose a random free port when using the --runbrowser option,
but stick to the fixed 8000 when using --startserver only.
------------------------------------------------------------------------
r36993 | fijal | 2007-01-19 14:54:25 +0100 (Fri, 19 Jan 2007) | 2 lines
Add possibility to scroll or not to scroll to web frontend. This checkin makes me believe that there is lot of stuff to do in JS backend.
------------------------------------------------------------------------
r36955 | hpk | 2007-01-18 17:44:21 +0100 (Thu, 18 Jan 2007) | 4 lines
some mods to defaultconfest dist options
and use "_dist_import_pypy", i.e. a leading underscore
at least
------------------------------------------------------------------------
r36943 | fijal | 2007-01-18 16:09:30 +0100 (Thu, 18 Jan 2007) | 3 lines
at least LSession should work now. Did not touched yet the remote options stuff,
but some simplifications were done.
------------------------------------------------------------------------
r36937 | fijal | 2007-01-18 14:22:54 +0100 (Thu, 18 Jan 2007) | 2 lines
Create a branch for further config cleanups.
------------------------------------------------------------------------
--HG--
branch : trunk
revisions 36936 to 37267:
------------------------------------------------------------------------
r37173 | fijal | 2007-01-23 11:26:32 +0100 (Tue, 23 Jan 2007) | 2 lines
Adhere to --box option.
------------------------------------------------------------------------
r37127 | fijal | 2007-01-22 13:48:33 +0100 (Mon, 22 Jan 2007) | 2 lines
Add a few docstrings.
------------------------------------------------------------------------
r37110 | hpk | 2007-01-22 00:39:24 +0100 (Mon, 22 Jan 2007) | 2 lines
renaming some variables for more clarity
------------------------------------------------------------------------
r37065 | arigo | 2007-01-20 17:15:10 +0100 (Sat, 20 Jan 2007) | 3 lines
Choose a random free port when using the --runbrowser option,
but stick to the fixed 8000 when using --startserver only.
------------------------------------------------------------------------
r37063 | arigo | 2007-01-20 17:00:31 +0100 (Sat, 20 Jan 2007) | 5 lines
webbrowser.open() may block until the browser finishes or not, so better
start it from a background thread. An case where it blocks is if you
define the $BROWSER environment variable.
------------------------------------------------------------------------
r37015 | fijal | 2007-01-19 17:24:32 +0100 (Fri, 19 Jan 2007) | 2 lines
Major cleanup of hostmanage. Got rid of global functions in favor of some better-structured classes.
------------------------------------------------------------------------
r36988 | fijal | 2007-01-19 13:25:05 +0100 (Fri, 19 Jan 2007) | 2 lines
Improve the C-c support, also keep web interface informed about crashes/interrupts.
------------------------------------------------------------------------
r36969 | fijal | 2007-01-18 20:10:36 +0100 (Thu, 18 Jan 2007) | 3 lines
Move teardown hosts out of finally (needs further refactoring)
Kill adddefaultoptions call for now.
------------------------------------------------------------------------
r36964 | hpk | 2007-01-18 19:30:02 +0100 (Thu, 18 Jan 2007) | 8 lines
move rconfig functionality to general Config object,
with docstrings, and trying to change rsession accordingly.
note that it's called "make_repr" and "merge_repr"
the latter to make it obvious that we are changing
the instance with the given repr.
------------------------------------------------------------------------
r36962 | fijal | 2007-01-18 19:06:13 +0100 (Thu, 18 Jan 2007) | 3 lines
Huge refactoring. Got rid of remote_options and session_options, two tests
still fail, so it's intermediate checkin.
------------------------------------------------------------------------
r36943 | fijal | 2007-01-18 16:09:30 +0100 (Thu, 18 Jan 2007) | 3 lines
at least LSession should work now. Did not touched yet the remote options stuff,
but some simplifications were done.
------------------------------------------------------------------------
r36937 | fijal | 2007-01-18 14:22:54 +0100 (Thu, 18 Jan 2007) | 2 lines
Create a branch for further config cleanups.
------------------------------------------------------------------------
--HG--
branch : trunk
revisions 36936 to 37267:
------------------------------------------------------------------------
r37178 | fijal | 2007-01-23 11:38:39 +0100 (Tue, 23 Jan 2007) | 2 lines
Make it a bit more working.
------------------------------------------------------------------------
r36937 | fijal | 2007-01-18 14:22:54 +0100 (Thu, 18 Jan 2007) | 2 lines
Create a branch for further config cleanups.
------------------------------------------------------------------------
--HG--
branch : trunk
revisions 36936 to 37267:
------------------------------------------------------------------------
r37263 | hpk | 2007-01-24 14:59:20 +0100 (Wed, 24 Jan 2007) | 19 lines
* introduce config.initdirect() which should
be the central place to perform initialization
on remote nodes/sides.
* fix --looponfailing (which works nicely now)
* have Conftest() be an instance instead of
a class-variable (avoid global state)
* more thoroughly use the "config.topdir" concept
(which is not neccessarily the pkgdir.dirpath())
* on merge_repr don't write values to
global defaultconftest module (this
gets tests failing that reuse defaultconftest
module, but also we want to avoid global
state in general)
------------------------------------------------------------------------
r37243 | hpk | 2007-01-24 00:06:24 +0100 (Wed, 24 Jan 2007) | 2 lines
small step-wise cleanups
------------------------------------------------------------------------
r37238 | hpk | 2007-01-23 23:09:57 +0100 (Tue, 23 Jan 2007) | 2 lines
why do it complicated?
------------------------------------------------------------------------
r37235 | hpk | 2007-01-23 22:25:16 +0100 (Tue, 23 Jan 2007) | 5 lines
added get_collector_trail method to get
a representation of a collector relative
to the topdir.
------------------------------------------------------------------------
r37230 | hpk | 2007-01-23 20:44:22 +0100 (Tue, 23 Jan 2007) | 2 lines
typo-level fixes
------------------------------------------------------------------------
r37193 | hpk | 2007-01-23 14:53:17 +0100 (Tue, 23 Jan 2007) | 5 lines
introduced config.topdir, tests and use it from terminal/remote at least.
also fix fijal's "--box" option handling a bit, but i am not
sure it should stay this way at all.
------------------------------------------------------------------------
r37172 | fijal | 2007-01-23 11:26:10 +0100 (Tue, 23 Jan 2007) | 2 lines
Boxing implies lsession
------------------------------------------------------------------------
r37114 | hpk | 2007-01-22 02:20:30 +0100 (Mon, 22 Jan 2007) | 6 lines
streamlining initial construction of collectors
(to unify topdir/pkgdir/collector handling
in rsession/terminal/normal)
still work in-progress with a skipped test
------------------------------------------------------------------------
r37112 | hpk | 2007-01-22 01:07:09 +0100 (Mon, 22 Jan 2007) | 3 lines
* moving towards a common rootdir/root collector
* adding an xxxed test
------------------------------------------------------------------------
r37110 | hpk | 2007-01-22 00:39:24 +0100 (Mon, 22 Jan 2007) | 2 lines
renaming some variables for more clarity
------------------------------------------------------------------------
r37077 | hpk | 2007-01-20 22:26:52 +0100 (Sat, 20 Jan 2007) | 2 lines
cleaning up fixoptions() a bit
------------------------------------------------------------------------
r37076 | hpk | 2007-01-20 22:19:42 +0100 (Sat, 20 Jan 2007) | 2 lines
unskip exec and looponfailing tests, fixes for that
------------------------------------------------------------------------
r37075 | hpk | 2007-01-20 21:51:18 +0100 (Sat, 20 Jan 2007) | 2 lines
remove two debug prints
------------------------------------------------------------------------
r37074 | hpk | 2007-01-20 21:43:52 +0100 (Sat, 20 Jan 2007) | 9 lines
* config.initsession() deals with getting at a session class
and initiliazing it
* config._getcollector(path) gets a collector pointing to path
* added lots of tests
* thus remove the strange getfscollector and map2colitems logic
* mark some tkinter tests as skipped
* fixing things all around
------------------------------------------------------------------------
r37065 | arigo | 2007-01-20 17:15:10 +0100 (Sat, 20 Jan 2007) | 3 lines
Choose a random free port when using the --runbrowser option,
but stick to the fixed 8000 when using --startserver only.
------------------------------------------------------------------------
r37061 | hpk | 2007-01-20 16:29:25 +0100 (Sat, 20 Jan 2007) | 8 lines
refactoring/fixing the remote terminal approach
(used for --exec and --looponfailing) to
use the new configiguration. Removes
lots of strange code.
XXX there are still some recursive testing issues.
------------------------------------------------------------------------
r37055 | hpk | 2007-01-20 12:51:54 +0100 (Sat, 20 Jan 2007) | 4 lines
reintroducing --session to lookup custom session objects in conftests.py
there is no other way to override a session object from conftests.
------------------------------------------------------------------------
r37025 | hpk | 2007-01-19 17:59:50 +0100 (Fri, 19 Jan 2007) | 2 lines
remove lget and its last usage (which was a bit random)
------------------------------------------------------------------------
r37020 | hpk | 2007-01-19 17:48:15 +0100 (Fri, 19 Jan 2007) | 3 lines
switching in the new way of getting to a sessionclass,
removing code in cmdline.py and elsewhere accordingly
------------------------------------------------------------------------
r37013 | hpk | 2007-01-19 17:14:11 +0100 (Fri, 19 Jan 2007) | 2 lines
somewhat simpler lookup mechanism for sessions
------------------------------------------------------------------------
r37009 | hpk | 2007-01-19 16:50:32 +0100 (Fri, 19 Jan 2007) | 2 lines
adding a config method to determine the session name
------------------------------------------------------------------------
r36994 | fijal | 2007-01-19 15:08:16 +0100 (Fri, 19 Jan 2007) | 2 lines
removal of some trailing white-spaces.
------------------------------------------------------------------------
r36964 | hpk | 2007-01-18 19:30:02 +0100 (Thu, 18 Jan 2007) | 8 lines
move rconfig functionality to general Config object,
with docstrings, and trying to change rsession accordingly.
note that it's called "make_repr" and "merge_repr"
the latter to make it obvious that we are changing
the instance with the given repr.
------------------------------------------------------------------------
r36942 | hpk | 2007-01-18 16:08:06 +0100 (Thu, 18 Jan 2007) | 4 lines
group things used for testing and not use non-underscore
names (otherwise they are consdiered public and
are possibly seen from apigen through py.test.config!)
------------------------------------------------------------------------
r36941 | fijal | 2007-01-18 15:59:08 +0100 (Thu, 18 Jan 2007) | 2 lines
Add a hack which allows to temporarily overwrite the config value.
------------------------------------------------------------------------
r36937 | fijal | 2007-01-18 14:22:54 +0100 (Thu, 18 Jan 2007) | 2 lines
Create a branch for further config cleanups.
------------------------------------------------------------------------
--HG--
branch : trunk
arigo