There’s a new release version of ROBOTC out now. No longer just a beta, this one’s for real! There are so many changes since the last release, 3.54, you’d have to be crazy not to update! I’ve been testing the in-between releases like mad, 3.56, 359 and some builds in between. They all seem to work fine with the Driver Suite. You’ll be happy to know that the latest version (3.3.1) of my Suite is included with this release.
This will become the minimum version of ROBOTC that the Driver Suite supports.
Something that may interest you is that we’re doing a live Google Hangout to chat about this release tomorrow (21 March 2013) at 14:00 (2PM) EST, that’s 19:00 CET (Amsterdam/Europe). Tim Friez, John Watson and myself will be there. So be there and be square! To get in on the action, just head to Robomatter on Google+ or visit ROBOTC.net/hangouts.
The text in red are changes between 3.59 and 3.60, the last beta. The other changes are from 3.54 to 3.59.
- New Standard Models Added for PLTW.
- Update VEX Cortex I2C interrupt handler to handle case where interrupt for last received character is not processed until STOP signal has already been sent.
- Add parameter to I2C Device Driver (VEX Cortex) to indicate whether being called for a “timer tick” or for a “message is now complete”. Add a few more statistics to the error collection.
- Increase VEX Cortex I2C message timeout to 6 milliseconds from 3.
- Start sending VEX Cortex I2C polling messages immediately after last poll has finished. Improves I2C responsiveness for VEX Cortex.
- Add watchdog timer support to VEX Cortex. This will allow the VEX Cortex User Processor to “restart” automatically once the processor stops responding. This is a useful workaround for any potential static issue that may cause the User Processor to lock up. Watchdog Timer is set to .75 seconds.
- Debugger “Motors” display – the last motor was not always properly updated. Fixed.
- Enable emulator for VEX PIC.
- Fix subsequent bug with “Motors” debugger display for VEX PIC locking up IDE when emulator is enabled; this bug should be limited to VEX PIC Emulator only.
- VEX PIC downloading was failing when master firmware was out of date. Getting stuck in a repetitive loop that wasn’t exiting. Fixed.
- VEX PIC was not executing user programs. Bug was that “start of flash file system VTOC” needed to be aligned on 4-byte boundary. Previous change had added a 1-byte field to the header preamble and there was not a corresponding 3-byte file added for VEX PIC. ARM platforms worked OK. This may also have broken Arduino (i.e. any 8-bit) platform).
- There are two separate flags for “allow any serial port for communications” — one for VEX Cortex and one for all other platforms. Preferences “Environment” tab was only updating flag for non-Cortex; this has been changed to update the appropriate flag based on platform setting. The VEX Cortex flag can also be updated in the “VEX Cortex” tab.
- User configurable UART setup was screwed up for the platforms whose “uart 0” is fixed as non-configurable and usable only for system communications port to PC — i.e. VEX Cortex, VEX PIC. The ROBOTC IDE was storing the data in the persistent data table (at start of flash file system) offset by one entry. I.E. data for “uart 1” on VEX Cortex was stored in “uart 2”, etc.
- Xander’s 3rd Party Driver Suite updated to version 3.3.1
- NXT color sensor read RGB individual values broken with pointer implementation.
- New feature for 3rd Party Sensor Driver Suite to set the two digital I/O lines on NXT sensors. Added two new sensor types (sensorCustom, sensorCustom9V) for this along with two new property intrinsic variables to set I/O line direction and values; these intrinsic variables are bit masks (2-bits) for the two lines.
- “setPixel” intrinsic (and corresponding “clearPixel” and “invertPixel”) take “unsigned” parameters. But implementation was using “signed” parameters and not properly range checking if parameters were negative. If negative, then should do nothing. Instead they were incorrectly wring to invalid buffer address which eventually caused a firmware crash.
- New firmware conditional compiles to eliminate the get/set property opcodes for ‘long’ and ‘float’.
- Arduino only. “Copy opcode from flash to RAM” performance improved.
- On Arduino, sensor initialization incorrectly assumed as sensor types had the extra “sensor buffer” structure. But the extra buffer is only linked for complicated sensor types so RAM was being overwritten. Fixed.
- Fix issue for “Motors” tab for Arduino in “Motors and Sensors Setup” property sheet. Was incorrectly trying to setup “encoder information for motor”, but Arduino platform does not “associate encoder sensor with motor” — which is currently only a feature for VEX Cortex.
- Updated application and desktop icon quality.
- Fix “Multiple overloads match” error introduced with change that did proper validation of pointer expression with point procedure parameter variable.
- Improved validation of procedure parameter matching for function overloads. Previously when parameter was a pointer, only checked to confirm that calling expression was also a pointer. Check was extended to also (correctly) match on the type of pointer as well.
- Could not change “check array bounds” in Compiler Code Optimization tab. The issue was that after changing the check mark the “code optimization type” would be recalculated and then the “optimization type” radio buttons updated. Array bounds checking is not used in determining the optimization type; but in setting the type it is modified. So the changed check mark was being accepted and then software was overwriting. Fixed.
- Split “auto save” into two individually settable flags: (1) to auto save source files before compile and (2) auto save files when application exits.
- Incorrect code optimization involving ‘struct’ variable offsets and perhaps other cases. For the assignment instruction “<variable2> is assigned address of <variable1>” compiler was replacing references to “<variable2> with “<variable1>. This optimization should only occur for “simple assignments” of “<variable2> = <variable1>”. Fixed.
- Change ‘nSysTime’ and ‘nPgmTime’ intrinsic variables to “unsigned”. ‘unsigned’ intrinsic variables were being converted to ‘signed’ when used in comparison expressions generating a warning regarding mixed ‘signed’ / ‘unsigned’ comparison. Fixed.
- Upper range check for “array bounds check” opcode used “>” instead of “>=” in comparison allowing ‘one past’ array size as valid.
- Added Support for Proxy Server when activating ROBOTC – Users can set Proxy preferences under “Detailed Preferences – General – Environment”.
- Incorrect generation of compiler error message when double pointer (i.e. “**”) is used.
- “File – Print” – a range of pages was not working when the starting page was not ‘1’. Fixed.
- Remove incorrect error message from error “Assignment between two different pointer types” when one side of ‘=’ is a ‘void *’.
- Compiler was incorrectly generating error message for “Expected a pointer expression in a pointer expression with ‘++’ or ‘-+’ operand. Compiler was incorrectly checking to verify the right operand was an integer value.
- Compiler was incorrectly generating error message for “Invalid pointer expression” in a pointer expression with ‘+’ or ‘-‘ operand. One operand of the expression must be a pointer and the other operand must be an integer (without implied conversion of a pointer to an integer). Compiler was incorrectly checking for the integer — the expression incorrectly had a “!” in it!
- Change code generation of ‘var arg’ for pointers to be consistent with standard C — i.e. the value of the pointer is placed in the argument. This may result in an additional instruction generated when argument is a ‘string’; previously RobotC got too cute trying to save this extra instruction which only worked for “%s” format codes but was broken for “%d” on a pointer.
- In ‘sprintf” implementation for “%s” and “%p” format codes– change from “get Address of parameter” to”get value of parameter” to match corresponding change in compiler code. Also minor cleanup of “%s” and “%p” format codes.
- Assignment expressions of “<char * variable> = <char constant> were generating compiler error because the “<char constant>” was evaluated as a “string” type. Added code in “get expression type” to check for this and not generate error as the compiler auto converts “string” to “char *” during code generation.
- Fixes an issue with the Live Start Page and “check for updates” that may have caused crashes.
- Improved validation of pointer expressions. Fix bug in calculating expression type of pointer expressions like “<ptr sub-expression> + <numeric constant>”; the result was (always?) incorrectly set as “long *”? Also added check that “=” of pointer is from a pointer of the same type – otherwise generate error message.
- Fixed consistency in implementation of “random” intrinsic property.
- Eliminate compiler error message in constant expression evaluation of sub-expressions using “*” operator. If either of the two operands is ‘zero’ then result is zero regardless of whether the other operand is a compile time constant.
- Do not generate additional errors — “too many parameters” specified — when procedure is “compiler generated undefined symbol”.
- When switch between “real robots” and “emulator”, the function that calculates “size of RAM pointer variable” was not being called. It adjusts between 2-byte (VEX PIC, Arduino) and 4-byte (Emulator, VEX Cortex, NXT) pointer sizes. No issue with VEX Cortex and NXT as they only use 4-byte pointers. But a problem with VEX PIC and Arduino where real robots use 2-byte RAM pointers! Added the appropriate call to function setup.
- Small code optimization for “postfix –“/postfix ++” operators to avoid temporary. In some cases they can be simplified to prefix operands.
- Incorrect initialization of static variables in “inner scope” for local procedure. They were initialized every time inner scope block was accessed rather than once on program startup.
- Debugger Panes for “Locals” and “Globals” (especially Globals) was not properly handling updates to ‘long’ and ‘float’ variables. Globals was completely broke — only lower two bytes of 4-byte variables was being updated which broke ‘long’ and ‘float’ variables. In both, ‘char’ variables were updating a short value — i.e. overwriting following characters.
- Fix problem with incorrect user code using a “short” variable and “sprintf” format code of “%f”. This can crash ROBOTC VM firmware if the short variable is not aligned on a 32-bit boundary.
- Add additional entries for StringFind for Character Constants and added test program.
- Redefine datalog opcodes and intrinsics. Legacy datalog incompatible with 3.5x VM operands which split memory variables and intrinsic properties into separate items.
- Firmware for all platforms now call “datalogHandlerInit()”. Conditional compile will define as NULL macro if a platform does not support Datalog. Datalog support is now conditionally compiled via “bHasDatalog” define rather than hard-coded as NXT only.
- Fix incorrect compiler type checking error when string constant is assigned to a char pointer.
- Fix bug in check for “is this a preprocessor string comparison expression”.
- Eliminate preprocessor string comparisons in Natural Language and replace with “defined(_Target_XXX_) where “_Target_XXX_” are three new system defined preprocessor variables — “_target_Robot_”, “_Target_Emulator_” and “Target_VirtWorld_”. Sample programs modified appropriately. Legacy user programs using legacy definitions will still work but will generate a compiler warning about non-standard extension.
- Add registry flag to enable compiler extension to allow preprocessor expressions support for string comparisons. Generate compiler error message if encountered without the flag being set.
- “cast” code generation fix. Previously when cast changed sign of result the ‘cast’ was applied before expression was converted to ‘int’ size used during a calculation. So casting a “ubyte” to “int” incorrectly converted it to a “signed char”. What should happen is “ubyte” expression gets evaluated into an “int” as part of expression evaluation (all expressions are evaluated at ‘int’ (or higher) level in “C”) and then the cast to “int” has no additional effect.
- Function to extract numeric “COMxxx”. Expanded syntax to support successful parsing of “(COM99)” previously would not accept extraneous characters.
- Add improved Dialog for selecting Communications Port. It uses a list box to display information about the port. Add check box to select any communications port.
- Compiler crash when parsing invalid syntax of ‘?’ expression. Compiler was not handling the case when “NULL” pointer returned from parsing sub-expression. The “NULL” was incorrect.