summaryrefslogtreecommitdiff
path: root/gpsfake.py
Commit message (Collapse)AuthorAgeFilesLines
* Tightening up the timing on the regressions a bit doesn't break them.Eric S. Raymond2009-01-211-2/+2
|
* Another try at defeating regression-timing bug.Eric S. Raymond2009-01-181-2/+2
| | | | | | | | | | This turns out to be at least a triple-point failure. Previous changes reduced the frequency and made it less random, but it appears gpsd has an unavoidable race condition when feeding through PTYs that can lose the last input packet. Workaround: compare test output with both the checkfile and the checkfile with its last line removed.
* Documentation fix.Eric S. Raymond2009-01-181-4/+1
|
* Forcing each pty to drain before closing appears to be helpful.Eric S. Raymond2009-01-161-0/+5
|
* Increasing the close timeout to 3 appears to banish timing errors.Eric S. Raymond2009-01-161-3/+3
|
* Make gpsfake use the C packet-getter...Eric S. Raymond2009-01-121-145/+28
| | | | | ...so it will no longer need to have its own duplicate knowledge of packet types wired in.
* Instrument gpsfake, in an attempt to track down an apparent bug in this module.Eric S. Raymond2009-01-121-8/+10
|
* More instrumentation.Eric S. Raymond2009-01-121-2/+5
|
* Beef up instrumentation a little.Eric S. Raymond2009-01-121-2/+2
|
* bzed's buffering fix makes it possible to cut the write padding a lot.Eric S. Raymond2009-01-121-1/+1
|
* Add some instrumentation.Eric S. Raymond2009-01-111-0/+3
|
* Remove a layer of code that was unecessary... Eric S. Raymond2009-01-101-3/+2
| | | | ...attacking the timing problems with the regression tests.
* Fixing broken identation as '\t' != ' 'Bernd Zeimetz2008-12-311-11/+11
|
* Ignore comments in NMEA test loads.Eric S. Raymond2008-12-311-1/+3
|
* I hade WRITE_PAD set too low, alas.Eric S. Raymond2008-12-311-1/+1
| | | | This caused the check logs to be regenerated incorrectly.
* Backing down to Python 2.4...Eric S. Raymond2008-12-301-1/+1
| | | | | ...to evade a hang bug in gpdfake, seems to require that WRITE_PAD be raised again.
* The fix from 15 Jul ("Bugs can lurk indefinitely in the damnedest places.")...Eric S. Raymond2008-12-301-1/+1
| | | | | ...makes it possible to run the test suite with much less padding than before.
* teach gpsfake how to play back itrax logs too...Chris Kuethe2007-12-281-0/+17
|
* This change will make gpsfake less fussy about trailing \n on binary logs.Eric S. Raymond2007-12-281-1/+3
| | | | It should enable the UBX log(s) to be read.
* An attempt at ubx support in gpsfake.Eric S. Raymond2007-12-281-0/+14
|
* String typo fix in bzed's code.Eric S. Raymond2007-12-201-1/+1
|
* Better path handling for gpsfake (patch by bzed).Eric S. Raymond2007-12-191-1/+14
|
* Sanity-check the Python modules with pychecker.Eric S. Raymond2007-12-141-16/+16
|
* Improvements to the test machinery, and another test.Eric S. Raymond2007-12-131-7/+10
|
* Portability tweaks.Eric S. Raymond2007-12-101-8/+11
|
* Avoid the regression hang on OpenBSD.Eric S. Raymond2007-12-101-11/+15
|
* Replace tcflush(..., TCIFLUSH) with ioctl(..., FIONREAD, ...)/read()...Eric S. Raymond2007-12-101-2/+11
| | | | | ...in the Python monitor class, as OpenBSD ptys apparently don't honor TCIFLUSH. The symptom is that the regression tests hang.
* Add more write padding to banish some fluky test failures.Eric S. Raymond2007-12-101-2/+6
|
* Diego Burge's driver builds, but it messes up some regression tests...Eric S. Raymond2006-12-151-0/+12
| | | | ...(notably the TSIP ones). There's a conflict...
* Add a new regression test, speed up all existing regression tests.Eric S. Raymond2006-12-121-1/+1
|
* This patch solve the OpenBSD TIOCDRAIN problem.Eric S. Raymond2006-12-071-4/+6
| | | | All regression tests pass under Linux.
* Change gpsfake to flush data written from gpsd before each feed. Eric S. Raymond2006-12-071-1/+8
| | | | | This may avoid hangs on OpenBSD; seems tcdrain() can hang until somebody actually reads the data on that system.
* Reintroduce line buffering in gps.py.Eric S. Raymond2006-12-061-1/+5
| | | | | Make gpsfake.py do realistic delays so that neither the pty buffer nor gpsd's own internal buffer gerts spammed.
* Eliminate threading from gpsfake.py.Eric S. Raymond2006-12-061-151/+128
| | | | | | | | | The TestSession class now schedules events itself. As a bonus, the new code is simpler and doesn't require system-dependent primitives that have to be implemented with /proc or fuser. There actually still is a threaded mode, untested, for use with valgrind-audit. But gpsfake doesn't need it.
* Checkpoint of the incomplete and half-tested support for setting serial...Eric S. Raymond2006-12-051-4/+36
| | | | | ...parameters in a test load. I'm checking it in ths way because it's time to try eliminating threads from gpsfake.py. The bugs are just too severe.
* Attempt to avoid an occasional race condition in gpsfake.Eric S. Raymond2006-12-051-0/+6
|
* Work around a potential race condition...Eric S. Raymond2006-12-051-3/+4
| | | | | ...when gpsfake tries to read a PID file that the daemon is not finished writing yet.
* Have gpsfake set VMIN the same way serial.c does.Eric S. Raymond2006-12-051-0/+1
| | | | All regression tests pass.
* Modified the gpsfake.py code to set serial-device modes...Eric S. Raymond2006-12-051-10/+22
| | | | | | | ...so that it does the same things as serial.c (in fact it's a line-by-line translation). All regression tests pass.
* Fix a comment.Eric S. Raymond2006-12-051-1/+1
|
* lsof is nasty. fstat is the way to go.Chris Kuethe2006-12-051-2/+2
|
* First cut at making gpsfake.py work using lsof(1) rather thanEric S. Raymond2006-12-021-12/+36
| | | | Linux-specific code.
* Make the -l option of gpsfake more useful.Eric S. Raymond2006-11-291-4/+9
|
* Fix a comment.Eric S. Raymond2006-11-291-2/+2
|
* gpsfake can now recognize and gather TSIP packets.Eric S. Raymond2006-11-291-0/+20
|
* Fix a minor bug: spurious "no probe matched" report.Eric S. Raymond2006-11-261-1/+1
| | | | Add a useful comment.
* Fix a long-standing bug in NMEA processing.Eric S. Raymond2006-11-241-1/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | My change to 'Fix a slightly broken test' (22 Nov 2006) unmasked a long-standing bug. GPGGA and GPGLL need to set the mode field when they have valid fix data, otherwise the O reports will often show no fix on a GLL or GGA that arrives before GSA in the send cycle (and some devices, like the FV-18, don't normally send GSA at all). This bug was being masked by an incorrect test; the have_fix() function was returning positive if status had been set to nonzero and no mode had yet been seen. This, in turn, caused some misbehavior in the P response; it would sometimes return P=? ?, which should not actually have happened. I spotted this symptom when looking at reports from Chris Kuethe's PHP script. This fix required rebuilding the regression tests, as the mode field on GLL and GGA sentences now frequently has a different (higher) value than it used to. It is quite possible that this bug accounted for a lot of the 2D/3D jitter that fans of the J patch were complaining about. Screwing with the mode field like this in the absence of received GSAs is a bit dodgy, but I don't see a better alternative. NMEA 0183 brain-damage strikes again -- what a %@!$&# botch of a protocol!
* Better recovery from a bad pid file.Eric S. Raymond2006-11-231-1/+6
|
* Add subversion ID tags near the beginning of every file.Chris Kuethe2006-09-211-0/+2
| | | | First line where possible, second line for scripts called with #!
* SiRF name cleanup.Chris Kuethe2006-08-171-1/+1
| | | | | | | | | | 1) the product line is called SiRFstar 2) I don't have any original SiRFstar receivers but my documentation says the protocol remained mostly the same between SS1, SS2 and SS3; it just got richer as time went by. At least with my SS2 and SS3 receivers, this holds true. 3) elsewhere in the code there are SIRF_THIS and SIRF_THAT, which implies (correctly) independence of chipset version.
View Lorry Controller Status