| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The routine which was determining if one path was a prefix of another
in arbitrating between explicit and generated resources was using the
raw string for the test without regard to path segments and thus could
be fooled by pairs such as "/tmp/foo" vs. "/tmp/foo2"
Fix was to be path delimiter aware and add a test.
Signed-off-by: Markus Roberts <Markus@reality.com>
|
| | | |
|
| | |
| | |
| | |
| | |
| | |
| | | |
Patch thanks to Till Maas
Signed-off-by: James Turnbull <james@lovedthanlost.net>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Replace deprecated method call. This code was not tested before, so I've
tried to capture what I think the method was trying to do.
This version includes Luke's suggested change to better preserve the
original behavior.
Signed-off-by: Jesse Wolfe <jes5199@gmail.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* use easily parsable format string (fixes #1872)
* only query stable packages, anything else cannot be installed anyway
* fix withenv usage in class method self.instances
* code cleanup & consistency
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This is a moderately ugly workaround for the MRI garbage collection
bug (see the ticket for details).
I explored several other potential solutions (notably, monkey
patching the routines that trigger the bug) but none of them were
satisfactory. Monkey patching sub, gsub, sub!, gsub!, etc., for
example, either changes the scoping of $~, $1, etc. in a way that
could potentially subtly change the meaning of programs or (if you
are clever) faithfully reproduces the behaviour of MRI--including
the memory leak.
I decided to go with the standardized and somewhat obnoxious never-
used optional argument as it was easy to automatically insert and
should be even easier to automatically find and remove if a better
fix is developed. It also should be obtrusive enough to escape
accidental removal in refactoring.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Puppet::Transaction was handling "tags" strings differently depending on
whether they came in from Puppet[:tags] or another source.
This was causing puppetrun's tags to be misparsed if there was not
exactly one --tag parameter.
I've moved the code to Util::Tagging.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Monkey-patch workaround for RDoc so it doesn't get confused by rubygems
stub executables.
Signed-off-by: Jesse Wolfe <jes5199@gmail.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This reverts commit c702f76b271515e9c42dcb923d379fbfac4c83cd and turns it
into a documentation only fix. As it turns out, ENV should have never been
used at all, as the Rack docs say nothing about it *and* Passenger's
behaviour in 2.2.3 + 2.2.4 was completely broken and still is half-way broken
in 2.2.5 (but is OK with the Rack specs).
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This modifies `supported_formats` to warn when the
`preferred_serialization_format` setting is invalid (and ignored in
favor of the default value).
I built the tests for this behavior alongside the existing
FormatHandler tests for *valid* `preferred_serialization_format` values
(and did some restructuring to extract common setup code to `before`
blocks).
Signed-off-by: Bruce Williams <bruce@codefluency.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Suprisingly, I found that setting allow_concurrency made the
"MySQL server has gone away" stop occuring even if the MySQL server
drops connections.
This may be the only change needed to restore compatibility with
ActiveRecord 2.1.x
Signed-off-by: Jesse Wolfe <jes5199@gmail.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
In #1469 we changed the FileSet internal api and missed that tidy
was still using the old version.
Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
|
| | | |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Puppetdoc wasn't using the current environment to find the correct
modulepath and manifestdir.
Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This adds a new environment cached attribute: manifestdir.
It is modeled on the code of modulepath, but returns
the manifestdir.
Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This makes parameters responsible for the canonicalization of their values and
provides a default (passthrough) implementation. It changes munge to pre-
canonicalize the value and resource references to builtin types to canonicalize
titles (which map to resorce namevars) with the corresponding parameter's
classes's canonicalization.
It adds a canonicalization routine to file paths that normalizes the behaviour
(trailing slashes are ignored) and DRYs up the related code.
Signed-off-by: Markus Roberts <Markus@reality.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
If setup code for a process depends on network connectivity
it needs to be protected with a rescue clause as much as the
main body of the process.
Further, Timeout exceptions aren't under StandardError and thus
aren't caught by an un-typed rescue clause. This doesn't matter
if we've morphed the exception, but will cause the program to
fail if we haven't.
There are many places where these concerns _might_ cause a problem
but in most cases they never will in practice; this patch addresses
the five cases where I have been able to confirm that it actually
can cause the client daemon to exit and two more where I suspect
(but can not prove) that it could.
This is an extension of the prior patch to cover additional cases
found by automated testing (repeated catalog runs with a 1% chance
of timeout forced on all timeout-bound operations, ~5000 runs).
The new cases recurred multiple times (>100 each) and in a final pass
with these corrected (~2500 runs) no additional cases were found.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Fix code that was passing an Array of code to a method that was
expecting a single tag.
Includes Markus's suggestions
Signed-off-by: Jesse Wolfe <jes5199@gmail.com>
|
| | | |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Ensure that resources whose refs are included in the catalog are
skipped to avoid duplication.
* Refactor to avoid early bailout on resources that cannot be ensured
absent.
* Remove check for managed? in generate
Checking if a resource is managed is unnecessary when checking for its
inclusion in the catalog.
* Add test coverage for Puppet::Type::Resources#generate
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The resource reference logic wasn't handling resources with "\n"s
in their namevars gracefully, and detection of this was complicated
by infelicitous exception reporting.
Note that this patch will require a merge when combined with the
patch for #2657.
Signed-off-by: Markus Roberts <Markus@reality.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Errors written to STDERR by Debian's invoke-rc.d are spurious, since
we're only using it to query whether a service can be started.
I'm adding the --quiet flag to suppress those messages.
Signed-off-by: Jesse Wolfe <jes5199@gmail.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This problem affects all types that generate sub-resources at
evaluation or pre-transaction time.
Thus it is fixed in the transaction, where we make sure we assign
all parent tags to the generated child resource.
A more correct fix would be to make sure we copy the whole context
(ie file, line, version, tags). This fix is planned for Rowlf.
Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
class klass {
include a, b, c
}
wasn't producing any rdoc documentation.
We were thinking code was always embedded in an array which is not
the case for mono-instruction code.
Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This patch fix this bug by adding more to_s methods to ast member
so that puppetdoc can just to_s the AST to reconstruct the original
puppet code.
Of course this is not perfect, but should work most of the time.
Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
When puppetdoc rdoc dies with an unhandled exception, the error
message produced didn't mention the file and line where the parser
choked.
This patch fixes this by trying to locate on which AST structure
the rdoc system failed.
Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This patch fixes a code-path in Puppet::Type::File::Content#insync? that
was failing to show a diff when the content's source came from a
template.
I was only able to reproduce one of the two such failures that Malcolm
Howe reported, but I think it's likely that they were both caused by
this same bug.
Signed-off-by: Jesse Wolfe <jes5199@gmail.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Since the storeconfig refactoring (ie moving the catalog storeconfig
system under the indirector) in 0.25 we lost the capability to
store the node ip and node environment name.
This patch restores this feature.
Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Since "preferred_serialization_format" was introduced, the older
"catalog_format" setting (yaml|marshal), when provided, has been
silently ignored and the default for "preferred_serialization_format"
used. This re-introduces "catalog_format," logs a deprecation notice,
and passes on any provided value to "preferred_serialization_format."
Signed-off-by: Bruce Williams <bruce@codefluency.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The fundemental problem was that, despite what the comment said,
the early bailout for file content management only applied to
directories, not to links. Making links bail out at as well fixed
the problem for most users.
However, it would still occur for users with mixed ruby version
system since there were no to_/from_pson methods for file metadata.
So the second (and far larger) part of this patch adds metadata
pson support.
The testing is unit level only, as there's no pratical way to do
the cross-ruby-version acceptance testing and no benifit to doing
"integration" testing short of that.
Signed-off-by: Markus Roberts <Markus@reductivelabs.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
It seems we never finalize the catalog, so the various default
resources are never created including the default schedules.
Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
|
| | |
| | |
| | |
| | |
| | |
| | | |
This is just Al's patch with removal of trailing ";"s.
Signed-off-by: Markus Roberts <Markus@reality.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
We now only warn when there's an actual change to make,
and we only make one warning per process run.
Signed-off-by: Luke Kanies <luke@madstop.com>
|
| | |
| | |
| | |
| | | |
Signed-off-by: Luke Kanies <luke@madstop.com>
|
| | | |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
been parsed
When fixing #2424, we were adding a global allow (ie allow(*)) to
the plugins/modules mount.
Unfortunately global allow always win against any other rules that
can be defined in fileserver.conf.
This patch makes sure we add those global allow entries only if
we didn't get any rules from fileserver.conf
Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
|
| | |
| | |
| | |
| | |
| | |
| | | |
This was deprecated in commit 1cfb0215 and was keeping puppetd from starting in listen mode.
Signed-off-by: John A. Barbuto <jbarbuto@corp.sourceforge.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
If setup code for a process depends on network connectivity
it needs to be protected with a rescue clause as much as the
main body of the process.
Further, Timeout exceptions aren't under StandardError and thus
aren't caught by an un-typed rescue clause. This doesn't matter
if we've morphed the exception, but will cause the program to
fail if we haven't.
There are many places where these concerns _might_ cause a problem
but in most cases they never will in practice; this patch addesses
the two cases where I have been able to confirm that it actually
can cause the client daemon to exit and two more where I suspect
(but can not prove) that it could.
I'd be willing to push this patch as it stands, as it at least
fixes demonstrable problems. A more general solution would be
nice.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This patch changes Puppet::Parser::Resource to check if it has been
passed two Puppet::Parser::Resource::Param objects with the same name.
Signed-off-by: Jesse Wolfe <jes5199@gmail.com>
|
| | |
| | |
| | |
| | |
| | |
| | | |
Change regexp to allow cron environment => "MAILTO="
Signed-off-by: Jesse Wolfe <jes5199@gmail.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This patches fixes a bug where setting an invalid option for "--color"
caused the word "nil" to be printed on every line of the log, instead
of printing out log messages.
Invalid color options now just produce uncolored output.
It seems to me that this isn't important enough to issue a warning
about an invalid setting.
Signed-off-by: Jesse Wolfe <jes5199@gmail.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Use of notify() is an error, so replace it with notice, which is a
function. Alternately, notify { msg => '2.6-1 is > than 2.4.5' } could
be used.
Signed-off-by: Todd Zullinger <tmz@pobox.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This affects several providers:
dpkg, rug, up2date, urpmi
Fixed by escaping the package names with Regexp.escape
Patch by Jesse Wolf
Signed-off-by: Markus Roberts <Markus@reality.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The latter is deprecated and was keeping puppetd from starting.
Also added a test and fixed a typo in another test.
Signed-off-by: John A. Barbuto <jbarbuto@corp.sourceforge.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
tags
f9516d introduced a change in the way the user tags are persisted
to the database: user tags are now treated as regular tags (they are
stored to the tags table).
Thus this commit changed the AR collector request to also look at the
tags tables when collecting.
Unfortunately this added a performance regression since tag request
were still importing the resources parameters tables and AR was
issuing a large request which was returning all the resource parameters
joined with the tags.
This commit fixes the AR request to join to the needed table, instead
of doing an include. Including (ie eager loading) parameter values was
not working for resource parameters anyway since at least 0.24 because
searching by parameter add a constraint to the joins and only the
searched parameter was returned instead of all parameter for a given
exported resource. So on a performance standpoint this new code should
be as fast 0.24 was.
Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
less often
This is not the right fix, but more a hackish workaround.
Since 0.25, the facts are transmitted as GET parameters when a
node asks for a catalog. Most proxies or webserver have a size limit
which is sometimes reached. In this case the request is denied
and the node can't get its catalog.
The idea is to compress facts (some non-scientific studies show a
57% fact size decrease for an average node) when transmitting
those when asking for a catalog.
Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
|
| | | |
|
|/ /
| |
| |
| | |
Signed-off-by: Markus Roberts <Markus@reality.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This commit extracts these three classes into a single
ResourceType class in the Parser heirarchy, now completely
independent of the AST heirarchy.
Most of the other changes are just changing the interface
to the new class, which is greatly simplified over the previous
classes.
This opens up the possibility of drastically simplifying a lot
of this other code, too -- in particular, replacing the reference
to the parser with a reference to the (soon to be renamed)
LoadedCode class.
Signed-off-by: Luke Kanies <luke@madstop.com>
|
| |
| |
| |
| | |
Signed-off-by: Luke Kanies <luke@madstop.com>
|