This adds a variable length dummy variable at the front of the EEPROM
when we erase it, which has the effect of moving the location of any
hot variables within the EEPROM. This should improve EEPROM
life. Thanks to Mike for the implementation.
Pair-Programmed-With: Mike Smith
git-svn-id: https://arducopter.googlecode.com/svn/trunk@3234 f9c3cf11-9bcb-44bc-f272-b75c42450872
when saving a variable, this avoids EEPROM wear by checking if the
existing value is already the same as the value being written, and
avoiding the write.
Thanks to Mike Smith for the implementation
git-svn-id: https://arducopter.googlecode.com/svn/trunk@3233 f9c3cf11-9bcb-44bc-f272-b75c42450872
We were using && instead of & for a bit check. Thanks to Mike for
spotting this!
Pair-Programmed-With: Mike Smith
git-svn-id: https://arducopter.googlecode.com/svn/trunk@3232 f9c3cf11-9bcb-44bc-f272-b75c42450872
This wipes all parameter values, and also clears out any waypoints as
the waypoint format could have changed between firmware revisions.
This also ensures that the AP_Var code can't re-use any key values
from the old EEPROM state, which makes us less dependent on the old
EEPROM being in a good state
Pair-Programmed-With: Mike Smith
git-svn-id: https://arducopter.googlecode.com/svn/trunk@3231 f9c3cf11-9bcb-44bc-f272-b75c42450872
This makes PSTR() type safe by using a 1 byte wrapper
structure. Attempts to use the wrong varient of a print function will
generate a compilation error.
git-svn-id: https://arducopter.googlecode.com/svn/trunk@1797 f9c3cf11-9bcb-44bc-f272-b75c42450872
Fix AP_Float16 - scaling was being done in the wrong direction, and the unserialiser wasn't reporting errors.
Add test cases for the above.
git-svn-id: https://arducopter.googlecode.com/svn/trunk@1696 f9c3cf11-9bcb-44bc-f272-b75c42450872
It might be possible to implement a set of virtual cast operator overrides, but only float lets us return NAN to indicate that a cast can't be made, so for now we'll stick with just this one.
git-svn-id: https://arducopter.googlecode.com/svn/trunk@1614 f9c3cf11-9bcb-44bc-f272-b75c42450872
Add pseudo-constant copies of the AP_Meta_calss::Type_id for the standard AP_VarT subclasses. This makes checking the type of a possible AP_Var much cheaper than calling ::meta_cast, as there is no object con/destruction involved.
Add some helper functions for manipulating AP_Vars from Menu interfaces.
git-svn-id: https://arducopter.googlecode.com/svn/trunk@1541 f9c3cf11-9bcb-44bc-f272-b75c42450872
Fix prototype for AP_Var_group::serialize so that it's called correctly when saving a group. Now group load/save works.
Add load/save_all and group load/save unit tests.
git-svn-id: https://arducopter.googlecode.com/svn/trunk@1533 f9c3cf11-9bcb-44bc-f272-b75c42450872
- We can't count on the first EEPROM scan knowing about everything; new variables may be constructed later so be willing to go back and re-scan the EEPROM if we encounter a new variable that hasn't been loaded yet.
- Sort out where _key actually points (the variable's data) in the EEPROM and get everyone to use the same interpretation.
- Fix return values from ::save and ::load.
- Make it possible to re-save after ::erase_all by de-locating all variables before the EEPROM is blown away.
- Fix args to eeprom_read_block/eeprom_write_block so that we don't spam random memory.
Add unit tests for basic save/load operations.
git-svn-id: https://arducopter.googlecode.com/svn/trunk@1532 f9c3cf11-9bcb-44bc-f272-b75c42450872
Now a variable belonging to a group can safely be constructed before the group it belongs to, and the group can be destroyed before its member variables. This greatly simplifies the AP_Var constructor(s).
Remove the lookup-by-index and lookup-by-key interfaces to AP_Var and replace them with first/next interfaces for all variables, and for variables belonging to a specific group. Document their usage. Add an accessor for the key associated with a variable so that search-by-key can be performed by a consumer. Throw away the lookup cache implementation, as it's not required anymore.
Re-layout the EEPROM variable header and tweak the EEPROM space allocator so that it's more resistant to interruptions during variable save.
Fix the global constants so that they work.
Add an interface for erasing all variables in EEPROM (only writes one byte).
Fix unit tests so that they work with the changed interfaces. Also tweak the unit test framework so that it doesn't inline all its code. This is a WIP - many more tests need to be written still.
git-svn-id: https://arducopter.googlecode.com/svn/trunk@1531 f9c3cf11-9bcb-44bc-f272-b75c42450872
Rather than implementing empty default constructors, make sure that AP_Var and friends have useful default constructors instead. This works around AP_Meta_class::meta_cast requiring a default constructor without having empty ctors scattered around.
Add accessors to AP_Var so that the global variable list can be traversed and the group/variable relationship comprehended by outsiders.
git-svn-id: https://arducopter.googlecode.com/svn/trunk@1523 f9c3cf11-9bcb-44bc-f272-b75c42450872
The format is a simple TLV format; a tag identifying the variable, a length for the variable, and then the actual data. This format could ultimately be extended to include waypoint/commands as well.
This allows us to load and save groups of variables to EEPROM without having to have a static EEPROM map.
Still needs unit tests and (probably) bugfixing.
git-svn-id: https://arducopter.googlecode.com/svn/trunk@1522 f9c3cf11-9bcb-44bc-f272-b75c42450872
AP_Vars with either a name or an address are 'interesting' (the latter so we can save_all).
Add the concept of address offsets to scopes. Now we have a container that we can put AP_Vars into that can be moved around in the EEPROM. This will make it easier for things like the PID library which need to support multiple instances getting their parameters from different parts of the ROM.
Improve documentation.
Suck it up and admit that we aren't going to do "identity"-based addressing for the EEPROM and just call the property "address".
git-svn-id: https://arducopter.googlecode.com/svn/trunk@1417 f9c3cf11-9bcb-44bc-f272-b75c42450872
The overriding principle here is to keep the use of AP_Vars as simple as possible, whilst letting the implementation do useful things behind the scenes. To that end, we define AP_Float, AP_Int8, AP_Int16 and AP_Int32. These are strongly typed, so that there is no ambiguity about what a variable "really" is.
The classes behave like the variables they are storing; you can use an AP_Float in most places you would use a regular float; you can add to it, multiply by it, etc. If it has been given an address in EEPROM you can load and save it.
Variables can be given names, and if they are named then they can be looked up. This allows e.g. a GCS or a test tool to find and traffic in variables that it may not explicitly know about.
AP_Var does not attempt to solve the problem of EEPROM address space management.
git-svn-id: https://arducopter.googlecode.com/svn/trunk@1399 f9c3cf11-9bcb-44bc-f272-b75c42450872
Andrew Tridgell
James Goppert
tridge60@gmail.com
james.goppert
DrZiplok
DrZiplok@gmail.com