aboutsummaryrefslogtreecommitdiff
path: root/debug
AgeCommit message (Collapse)AuthorFilesLines
2017-03-03Print an error message when the OpenOCD config file can't be readPalmer Dabbelt1-1/+6
2017-02-24Add assert to save me from myself.Tim Newsome2-4/+5
spike passes the download test, but needs a 15 minute timeout. That's too ridiculous to put in here, but 60s is reasonable.
2017-02-24Keep the DownloadTest source if the test fails.Tim Newsome1-14/+17
2017-02-24Make pylint happy.Tim Newsome1-3/+4
2017-02-21Talk to spike using OpenOCD instead of directly.Tim Newsome2-13/+27
2017-02-17Add HiFive1 target.Tim Newsome3-1/+63
2017-01-02Increase delay in UserInterrupt.Tim Newsome1-1/+1
This makes the test pass reliably (10/10) against the Arty board. Previously it failed intermittently.
2016-12-27Use compressed code if the target supports it.Tim Newsome6-61/+109
The main change was to read misa before running any other test. If misa indicates C is supported, then use compressed code. This required changing some tests, mostly to ensure correct alignment. The single step test also needs to know the correct addresses to step through in compressed code. Only print at most 1000 lines from each log file.
2016-12-27Clean up .pyc files.Tim Newsome1-1/+1
2016-12-13Fix WriteCsrs test.Tim Newsome1-2/+2
At some point the program changed to use a different register and this test was never updated. If $x1 pointed somewhere bad, that would result in an exception.
2016-12-12Pass newly updated -march, -mabi options to gccAndrew Waterman1-2/+2
2016-12-12Disable tcl and telnet servers when running OpenOCD because the port numbers ↵Richard Xia1-2/+12
might conflict with other processes.
2016-12-09Add test of compare-sections command.Tim Newsome1-0/+10
2016-12-07Use XLEN macro for these sources as well.Tim Newsome3-4/+4
All tests pass on spike32 and spike64 again.
2016-12-07Use our own XLEN macro.Tim Newsome2-1/+2
Relying on something that the compiler automatically sets is apparently not reliable.
2016-12-07We *do* need the FPU to compile 64-bit code.Tim Newsome1-0/+1
2016-12-07Don't compile with FPU support for now.Tim Newsome1-1/+5
It's not Just Working, and none of the tests so far actually care.
2016-12-07Nicely display compile failures.Tim Newsome1-8/+20
2016-12-07Don't eat compile errors.Tim Newsome1-12/+3
2016-12-07Fix race when finding the port OpenOCD is using.Tim Newsome1-6/+11
2016-12-06avoid non-standard predefined macrosAndrew Waterman4-6/+6
2016-11-14Attempt to work around hard-float linking problemAndrew Waterman1-1/+1
2016-11-10Decrease sleep time to 0.1s.Richard Xia1-2/+2
2016-11-04Tell OpenOCD to pick an unused port, and use lsof to figure out which one it ↵Richard Xia1-3/+27
picked.
2016-10-31Add basic floating point register test.Tim Newsome3-3/+31
2016-10-21Improve reg test a little.Tim Newsome1-6/+14
2016-10-20Test OpenOCD step and resume.Tim Newsome2-6/+56
2016-10-19Fix conflict: JTAG VPI changes vs openocd refactorTim Newsome1-7/+4
2016-10-19Actually test something in regs test.Tim Newsome1-2/+4
2016-10-19Merge pull request #34 from richardxia/use-port-randomizationTim Newsome3-4/+18
Use port randomization for VCS and OpenOCD
2016-10-18Pull port number from VCS output and pass to OpenOCD.Richard Xia3-2/+10
2016-10-18Add framework to test OpenOCD directly.Tim Newsome4-298/+405
This took a lot of refactoring to make it look reasonable. There isn't actually any functional OpenOCD test yet. But a dummy test runs a command (and fails).
2016-10-18Randomize gdb port.Richard Xia1-2/+8
2016-10-14Wait for OpenOCD to start fully before connecting.Tim Newsome1-0/+15
Requires OpenOCD change 4eba841bfee4b6c347423ac28851ab7ee7e75532
2016-10-13Add MemTestWriteInvalid.Tim Newsome1-1/+13
2016-10-03Add test for memory read from invalid address.Tim Newsome2-0/+12
2016-09-30Tolerate remotes that return memory read errors.Tim Newsome1-1/+9
2016-09-29Update dmode test to match spec.Tim Newsome2-46/+101
M-mode writes to triggers with dmode set are now ignored instead of raising an exception. Also added -f/--fail-fast option to gdbserver.
2016-09-29Clear triggers during entry.Tim Newsome1-0/+9
If the last test leaves some triggers set they should be cleaned up.
2016-09-28Disable TriggerDmode while spike is changed.Tim Newsome1-0/+4
2016-09-20Minor tweaks to trigger tests for openocd.Tim Newsome2-10/+16
2016-09-15Rewrite debug testing.Tim Newsome4-264/+762
No longer use unittest. Now tests can return not_applicable if eg. a desired execution mode isn't implemented on a target. Also we do a better job killing spike processes when a test fails. Did a lot of code cleanup, partly by using pylint. Fix the Makefile so that if the test fails, 'make' actually fails too.
2016-09-02Fix/add to instant trigger tests.Tim Newsome2-6/+23
2016-09-02Limit spike RAM so I can run valgrind on it.Tim Newsome1-0/+1
2016-09-01Add some immediate trigger tests.Tim Newsome2-6/+23
2016-09-01Create TriggerTest.Tim Newsome2-21/+136
2016-09-01Add test for address triggers.Tim Newsome2-1/+29
2016-09-01Remove "import ." notation.Tim Newsome1-1/+1
Doesn't work with Python 2.7.9 which is the default in latest Debian.
2016-08-23Use a version range.Richard Xia1-1/+1
2016-08-22Add requirements.txt and reorder imports by type.Richard Xia3-4/+6