On 10/30/14 11:11, Isaac Dunham wrote: > Test ls -d to make sure it does not follow links.
I looked at that this morning but didn't update it because some of the existing tests are failing, so I have to go through the whole thing and poke at it, and right now I'm doing sed.tests. :) (I don't have a "pending" mechanism for tests. I suspect doing a proper pedantic thorough cleanup and fill-out pass on the test suite is going to be either right before 1.0 or right after it.) Looking closer at ls.tests, the "ls -a" test is failing because the dot entries are at the end instead of the beginning. Except that behavior is locale dependent at the best of times, if I do LC_ALL=c TEST_HOST=1 scripts/test.sh ls Now the _host_ ls is failing the tests. Probaby we want to run the output through "sort" to get stable behavior out of it. Also there are 26 command line options (not counting --color), and 16 tests in testing. So what's there is unlikely to be all that thorough... Proper test suite entries can eat an _amazing_ amount of time if you let them. The possible pedantic nitpicking is just amazing, a _proper_ test suite will do "ls doesnotexit 2>/dev/null || echo err" and "ls > /dev/full 2>/dev/null || echo err" and so on. Not just testing what should work but what _shouldn't_, testing that failures are caught and errors are reported. Toybox needs to be simple, needs to be compact, needs to implement all the appropriate functionality, needs to be robust about catching failures and reporting errors, needs to have good help text for each command explaining what it does and how to use it, needs error messages that don't expect the reader to know more that a couple dozen words of english (about the level C keywords or posix command names already require), needs its internals well documented, needs a thorough test suite, and this is before you get into making utf8 work and supporting nommu systems and bootloader embedding and how we don't want to rely on any external libraries but _would_ like to support https and there's still no decent embedded stunnel implementation around... If you wonder why it's taking so much time, there's an awful lot to do. Thanks for the update though. :) > -- > On Thu, Oct 30, 2014 at 05:37:35AM -0500, Rob Landley wrote: >> Got it. >> http://landley.net/hg/toybox/rev/1540 >> Thanks, >> Rob > > Also, looking at man 1p ls, it looks like -HL takes priority over -dFl. > (Yes, -F also implies not following links.) By my reading of the spec, -HL is supposed to take priority? I checked in adding -F. > Thanks, > Isaac Dunham Thanks, Rob _______________________________________________ Toybox mailing list [email protected] http://lists.landley.net/listinfo.cgi/toybox-landley.net
