Fix MarkDown syntax
This commit is contained in:
parent
7b87b061da
commit
63ddc0f6ba
2 changed files with 45 additions and 29 deletions
|
@ -1,17 +1,17 @@
|
|||
= CMake in FlightGear overview =
|
||||
# CMake in FlightGear overview
|
||||
|
||||
CMake has evolved considerably over the years; if you're reading external tutorials on it, ensure
|
||||
they mention 'modern CMake' since it's very different to the older style.
|
||||
CMake has evolved considerably in the past decade; if you're reading external tutorials abput it, ensure
|
||||
they mention 'modern CMake', or the information will be incorrect.
|
||||
|
||||
The top-level CMakeList.txt handles configuration options, finding dependencies, and scanning
|
||||
the rest of the source tree. Across the source tree we add executables, includ the main FGFS
|
||||
binary but also other helpers and utilities. We also define targets for helper libraries for
|
||||
The top-level `CMakeLists.txt` handles configuration options, finding dependencies, and scanning
|
||||
the rest of the source tree. Across the source tree we add executables, including the main FGFS
|
||||
binary but also other helpers and utilities. We also define targets for helper libraries, for
|
||||
various reasons; for example to build some code with different include paths or flags.
|
||||
|
||||
Due to the original code structure, we use some helper functions to collect most of the
|
||||
Due to the historical code structure, we use some helper functions to collect most of the
|
||||
application sources into two global CMake variables, which are then read and added to the
|
||||
main executable target, in `src/Main/CMakeList.txt`. Therefore, many subdirectories have
|
||||
a trivial CMakeList.txt which simply calls the helper functions to add its sources:
|
||||
a trivial `CMakeLists.txt` which simply calls the helper functions to add its sources:
|
||||
|
||||
```
|
||||
include(FlightGearComponent)
|
||||
|
@ -29,25 +29,30 @@ set(HEADERS
|
|||
flightgear_component(MyComp "${SOURCES}" "${HEADERS}")
|
||||
```
|
||||
|
||||
== Configurations ==
|
||||
The global properties used are `FG_SOURCES` and `FG_HEADERS`.
|
||||
|
||||
Release builds are built with RelWithDebInfo; this is also the most useful configuration for
|
||||
development, sicne on Windows, Debug is unusably slow. If trying to optimise performance,
|
||||
## Configurations
|
||||
|
||||
Official release builds are built with `RelWithDebInfo`; this is also the most useful configuration for
|
||||
development, since on Windows, `Debug` is unusably slow. If trying to optimise performance,
|
||||
keep in mind that compiler flags must be manually set for `RelWithDebInfo`; they are _not_
|
||||
automatically inherited from `Release`.
|
||||
|
||||
Adding additional configurations is possible: for example for profiling. CMake also picks up
|
||||
the `CXXFLAGS` environment variable to pass ad-hoc compiler options, without modifying the
|
||||
build systen.
|
||||
|
||||
== Dependencies ==
|
||||
## Dependencies
|
||||
|
||||
All dependencies should be handled via an `IMPORTED` target: this ensures that include paths,
|
||||
link options, etc specific to the dependency are handled correclty across different platforms.
|
||||
link options, etc specific to the dependency are handled correctly across different platforms.
|
||||
|
||||
Depending on the dependency, it may supply a Foo-Config.cmake which defines such a target for
|
||||
For some dependencies, there may be a zFoo-Config.cmakez which defines such a target for
|
||||
you automatically. Or there may be an existing `FindFoo.cmake` which does the same. If neither
|
||||
of these situations exist, create a custom finder file in `CMakeModules`, following the existing
|
||||
examples.
|
||||
|
||||
CMake tracks transitive dependencies correctly, so for example if your new dependency is used
|
||||
CMake tracks transitive dependencies precisely, so for example if your new dependency is used
|
||||
in SimGear, it will automatically be added to the include / link paths for FlightGear based on
|
||||
the SimGear build type.
|
||||
|
||||
|
|
|
@ -1,13 +1,13 @@
|
|||
= Sentry.io Cash Reporting =
|
||||
# Sentry.io Cash Reporting
|
||||
|
||||
FlightGear can optionally report crashes and serious errors to Sentry.io. We have a sponsored
|
||||
account provided gratis by Sentry; for access to this, ask on the developer list.
|
||||
|
||||
== Error conditions ==
|
||||
## Error conditions
|
||||
|
||||
If FlightGear crahses, SDentry will automatically submit a report. For non-crash errors,
|
||||
If FlightGear crahses, Sentry will automatically submit a report. For non-crash errors,
|
||||
we manually submit an exception report. At present this is done whenever we trigger the
|
||||
'fatalMessageBox', and in other serious situations. Deciding where is appropriate (or not)
|
||||
`fatalMessageBox`, and in other serious situations. Deciding where is appropriate (or not)
|
||||
to report the error to Sentry is a key challenge of the system, since we don't want to
|
||||
report user misconfiguration problems, but we do want to detect recurring and
|
||||
systematic failures, eg broken aircraft.
|
||||
|
@ -17,17 +17,19 @@ better to send errors and filter them on the server side, but this needs to be d
|
|||
with some intelligence; for example we do not (at present) report Nasal runtime
|
||||
errors, since this might overload the system.
|
||||
|
||||
== Data Protection ==
|
||||
## Data Protection
|
||||
|
||||
We explicitly don't include any personal information in the reports, and disable IP address
|
||||
collection. Since this makes it hard to cluster reports by user, we instead generate a UUID
|
||||
We explicitly do not include any personal information in the reports, and disable IP address
|
||||
collection. This avoids any GPDR obligations for us.
|
||||
Since this makes it hard to cluster reports by user, we instead generate a UUID
|
||||
corresponding to a FlightGear installation. This gives us a way to anonymously cluster reports
|
||||
by user, without any personal data disclosure.
|
||||
by computer, without any personal data disclosure. (So we can determine if a hundred reports of a
|
||||
crash come from fifty discrete users, or just one)
|
||||
|
||||
== Supplementatal Data ==
|
||||
## Supplementatal Data
|
||||
|
||||
We record various pieces of configuration state as 'tags': such as the OS, graphics card,
|
||||
key settings (eg, is multi-player enabled, which aircraft is being flown) as _tags_ in
|
||||
major settings (eg, is multi-player enabled, which aircraft is being flown) as _tags_ in
|
||||
Sentry terminology. This allows grouping of reports by similar tags; for example we can
|
||||
identify that a particular crash only occurs with Intel Graphics, or when flying the
|
||||
C172.
|
||||
|
@ -37,16 +39,25 @@ is sent, and give an idea of the path of the user through the application, prior
|
|||
crash. We have breadcrumbs for key events such as the splash-screen completing, the
|
||||
user changing position, or scenery being reloaded.
|
||||
|
||||
WARN and ALERT level `SG_LOG` messages are currently included as breadcrumbs automatically;
|
||||
this means it's important not casually add messages at the levels for non-serious conditions.
|
||||
`WARN` and `ALERT` level `SG_LOG` messages are currently included as breadcrumbs automatically;
|
||||
this means it's important not to casually add messages at the levels for non-serious conditions.
|
||||
The integration code has a list of commonly ocurring but non-useful messages which are
|
||||
skipped from sending; especially some OSG ones related to minor PNG and AC3D data issues.
|
||||
skipped from sending; especially some OSG ones related to PNG and AC3D data issues.
|
||||
|
||||
Adding new tags or breadcrumbs should be done with care, but is generally useful, and
|
||||
suggestions in this area are appreciated.
|
||||
|
||||
== API ==
|
||||
## API
|
||||
|
||||
All the API is contained in `sentryIntegration.hxx`, inside the `flightgear` namespace.
|
||||
The methods are no-ops if the Sentry SDK was not available at CMake time, and are also
|
||||
no-ops if the user has disabled sentry reporting.
|
||||
|
||||
## Building
|
||||
|
||||
The API requires the injection of a private API key to report to our account; this
|
||||
must be kept private of course, so it's injected into official builds at CMake time
|
||||
on Jenkins, via the environment variable `FLIGHTGEAR_SENTRY_API_KEY`.
|
||||
|
||||
The build must be configured to produce debug symbols, which are uploaded to Sentry
|
||||
via the `sentry-cli` tool as part of our build scripts.
|
Loading…
Reference in a new issue