On Friday 22 February 2008, Francis Moreau wrote: > > > > > > To start the tests, I use the testusb user tool with the wrapper script > > test.sh. > > > In this script, the buflen for the host is set to 2048. > > > > Which as you note is incompatible with "pattern=1". > > > > If you want to use "pattern=1" you need a different test script. > > Well we could do a set_interface of set_config after each loop that way the > pattern restart to 0.
That's done *between tests* instead ... if you really want to have very tiny tests, you can already do that, no changes needed (except, as noted, to the test script). But that would make for poor stress tests. The couldn't trigger certain types of failures (think: races) that happen while the peripheral side test driver (g_zero) is "between buffers" yet the host is still throwing data at the hardware. There are a lot of tricky driver and controller states there, which a test framework needs to have a fair chance at triggering. > Or make the pattern %64 instead of %63 could work when buflen is a power > of 2. The point of using a modulo-63 counter is to ensure that certain errors get detected ... like duplication of a 64-byte full speed bulk packet, as I've seen be caused by both software and hardware bugs. A modulo-64 would be pretty useless at detecting such bugs. - Dave - To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
