hg-loggingmod 0.1.3

Managing Mercurial logs with the standard Python logging module

Mercurial logs through the logging module

This extension redirects most user feedback meant for terminals to the Python logging module of the standard library.

This is intended intended primarily for server-side use cases, and is unlikely to be useful for client-side operation.

Indeed on a server, many messages are useful for the systems administrator only, and would be unwanted pollution if seen by the client. A notable exception is ui.status which is really used to report back meaningful information over the wire.

Using logging is more robust and flexible than using flags such as ui.debug and redirecting stderr. It also provides integration with Sentry (see below). Many other logs aggregation systems that have a logging handler could be directly used.

At the time of this writing, it has the side effect to disable other extensions meant for logs, such as blackbox.

Installation

WARNING: this is completely untested at this point with Python 3, expected not to really work.

Install from the source tree using pip install .. Then, in your HGRC, include this:

[extensions]
loggingmod =

Configuration

All parameters are to be set within the [logging] hgrc section.

Basic configuration

These are applied first, using mostly logging.basicConfig

Example (these are the default values):

[logging]
level = info
format = [%(asctime)s] [%(process)d] [%(levelname)s] [%(name)s] %(message)s
hg_format = [%(asctime)s] [%(process)d] repo:%(repo)s [%(levelname)s] [%(name)s] %(message)s

Logging to a file

[logging]
file = /var/log/mercurial.log

Format

The format string is a regular logging format string, see LogRecord attributes for a full list of the keys that can be used there.

The hg_format string is also a regular logging format string, but it can additionally make use of the repo parameter: the full path to the current repository on the file system, if relevant to the given log record. This format is used in the hg logger and its descendants, such as hg.logging, hg.discovery etc.

The times will include the timezone by default.

Advanced configuration through files

JSON

This is the most complete, as it leverages dictConfig()

[logging]
config.json = /etc/hg-logging.json

The basic configuration above is done first nevertheless, but it is wiped out unless one uses the incremental keyword.

INI

This is forwarded to fileConfig()

[logging]
config.ini = /etc/hg-logging.ini

Using with Sentry

Prerequisites

The sentry_sdk package should be importable from the Mercurial process, see Sentry install instructions

In our experience, pip install sentry-sdk has not been enough, we had to also install Brotli as well. Your mileage may vary.

In doubt, test it first

Activation

To forward logs to Sentry, just specify the DSN in the hgrc:

[logging]
sentry.dsn = https://<key>@sentry.example.net/<project>

Warning : don't use quotes above.

Basic configuration

The Sentry default integration catches all logging calls, and is orthogonal to the regular logging configuration, except for the logger levels.

This extension has several knobs to tweak it. Here's an example

[logging]
sentry.ignore_loggers = discovery extension

# these are the default values:
sentry.event_level = error
sentry.breadcrumb_level = info

Fine configuration

Instead of the flat configuration as above, you can disable the blanket integration and resort to explicit configuration with the config.json directive and Sentry's handlers

[logging]
config.json = /etc/hg-config-with-sentry.conf
sentry.dsn = https://<key>@sentry.example.net/<project>
sentry.default_integration = false

Of course, this also disables the sentry.event_level and sentry.breadcrumbs_level config items.

At the time of this writing, these handlers are

• sentry_sdk.integrations.logging.EventHandler
• sentry_sdk.integrations.logging.BreadCrumbHandler

See also: Handler classes in Sentry documentation