AnyEvent::Debug - debugging utilities for AnyEvent
use AnyEvent::Debug;
# create an interactive shell into the program
my $shell = AnyEvent::Debug::shell "unix/", "/home/schmorp/myshell";
# then on the shell: "socat readline /home/schmorp/myshell"
This module provides functionality hopefully useful for debugging.
At the moment, "only" an interactive shell is implemented. This shell
allows you to interactively "telnet into" your program and execute
Perl code, e.g. to look at global variables.
- $shell = AnyEvent::Debug::shell $host, $service
- This function binds on the given host and service port and
returns a shell object, which determines the lifetime of the shell. Any
number of connections are accepted on the port, and they will give you a
very primitive shell that simply executes every line you enter.
All commands will be executed "blockingly" with the socket
"select"ed for output. For a less "blocking" interface
see Coro::Debug.
The commands will be executed in the "AnyEvent::Debug::shell"
package, which currently has "help" and a few other commands,
and can be freely modified by all shells. Code is evaluated under
"use strict 'subs'".
Every shell has a logging context ($LOGGER) that is attached to
$AnyEvent::Log::COLLECT), which is especially useful to gether debug and
trace messages.
As a general programming guide, consider the beneficial aspects of using
more global ("our") variables than local ones ("my")
in package scope: Earlier all my modules tended to hide internal variables
inside "my" variables, so users couldn't accidentally access
them. Having interactive access to your programs changed that: having
internal variables still in the global scope means you can debug them
easier.
As no authentication is done, in most cases it is best not to use a TCP
port, but a unix domain socket, which can be put wherever you can access
it, but not others:
our $SHELL = AnyEvent::Debug::shell "unix/", "/home/schmorp/shell";
Then you can use a tool to connect to the shell, such as the ever versatile
"socat", which in addition can give you readline support:
socat readline /home/schmorp/shell
# or:
cd /home/schmorp; socat readline unix:shell
Socat can even give you a persistent history:
socat readline,history=.anyevent-history unix:shell
Binding on 127.0.0.1 (or "::1") might be a less secure but still
not totally insecure (on single-user machines) alternative to let you use
other tools, such as telnet:
our $SHELL = AnyEvent::Debug::shell "127.1", "1357";
And then:
telnet localhost 1357
- AnyEvent::Debug::wrap [$level]
- Sets the instrumenting/wrapping level of all watchers that
are being created after this call. If no $level has been specified, then
it toggles between 0 and 1.
The default wrap level is 0, or whatever $ENV{PERL_ANYEVENT_DEBUG_WRAP}
specifies.
A level of 0 disables wrapping, i.e. AnyEvent works normally, and in its
most efficient mode.
A level of 1 or higher enables wrapping, which replaces all watchers by
AnyEvent::Debug::Wrapped objects, stores the location where a watcher was
created and wraps the callback to log all invocations at "trace"
loglevel if tracing is enabled fore the watcher. The initial state of
tracing when creating a watcher is taken from the global variable
$AnyEvent:Debug::TRACE. The default value of that variable is 1, but it
can make sense to set it to 0 and then do "local
$AnyEvent::Debug::TRACE = 1" in a block where you create
"interesting" watchers. Tracing can also be enabled and disabled
later by calling the watcher's "trace" method.
The wrapper will also count how many times the callback was invoked and will
record up to ten runtime errors with corresponding backtraces. It will
also log runtime errors at "error" loglevel.
To see the trace messages, you can invoke your program with
"PERL_ANYEVENT_VERBOSE=9", or you can use AnyEvent::Log to
divert the trace messages in any way you like (the EXAMPLES section in
AnyEvent::Log has some examples).
A level of 2 does everything that level 1 does, but also stores a full
backtrace of the location the watcher was created, which slows down
watcher creation considerably.
Every wrapped watcher will be linked into %AnyEvent::Debug::Wrapped, with
its address as key. The "wl" command in the debug shell can be
used to list watchers.
Instrumenting can increase the size of each watcher multiple times, and,
especially when backtraces are involved, also slows down watcher creation
a lot.
Also, enabling and disabling instrumentation will not recover the full
performance that you had before wrapping (the AE::xxx functions will stay
slower, for example).
If you are developing your program, also consider using AnyEvent::Strict to
check for common mistakes.
- AnyEvent::Debug::path2mod $path
- Tries to replace a path (e.g. the file name returned by
caller) by a module name. Returns the path unchanged if it fails.
Example:
print AnyEvent::Debug::path2mod "/usr/lib/perl5/AnyEvent/Debug.pm";
# might print "AnyEvent::Debug"
- AnyEvent::Debug::cb2str $cb
- Using various gambits, tries to convert a callback (e.g. a
code reference) into a more useful string.
Very useful if you debug a program and have some callback, but you want to
know where in the program the callback is actually defined.
- AnyEvent::Debug::backtrace [$skip]
- Creates a backtrace (actually an AnyEvent::Debug::Backtrace
object that you can stringify), not unlike the Carp module would. Unlike
the Carp module it resolves some references (such as callbacks) to more
user-friendly strings, has a more succinct output format and most
importantly: doesn't leak memory like hell.
The reason it creates an object is to save time, as formatting can be done
at a later time. Still, creating a backtrace is a relatively slow
operation.
All watchers created while the wrap level is non-zero will be wrapped inside an
AnyEvent::Debug::Wrapped object. The address of the wrapped watcher will
become its ID - every watcher will be stored in
$AnyEvent::Debug::Wrapped{$id}.
These wrapper objects can be stringified and have some methods defined on them.
For debugging, of course, it can be helpful to look into these objects, which is
why this is documented here, but this might change at any time in future
versions.
Each object is a relatively standard hash with the following members:
type => name of the method used to create the watcher (e.g. C<io>, C<timer>).
w => the actual watcher
rfile => reference to the filename of the file the watcher was created in
line => line number where it was created
sub => function name (or a special string) which created the watcher
cur => if created inside another watcher callback, this is the string rep of the other watcher
now => the timestamp (AE::now) when the watcher was created
arg => the arguments used to create the watcher (sans C<cb>)
cb => the original callback used to create the watcher
called => the number of times the callback was called
Each object supports the following mehtods (warning: these are only available on
wrapped watchers, so are best for interactive use via the debug shell).
- $w->id
- Returns the numerical id of the watcher, as used in the
debug shell.
- $w->verbose
- Returns a multiline textual description of the watcher,
including the first ten exceptions caught while executing the
callback.
- $w->trace ($on)
- Enables ($on is true) or disables ($on is false) tracing on
this watcher.
To get tracing messages, both the global logging settings must have trace
messages enabled for the context "AnyEvent::Debug" and tracing
must be enabled for the wrapped watcher.
To enable trace messages globally, the simplest way is to start the program
with "PERL_ANYEVENT_VERBOSE=9" in the environment.
Tracing for each individual watcher is enabled by default (unless
$AnyEvent::Debug::TRACE has been set to false).
Marc Lehmann <[email protected]>
http://anyevent.schmorp.de