On Mon, Jun 9, 2014 at 3:13 PM, Ben Pfaff <b...@nicira.com> wrote: > On Fri, May 30, 2014 at 12:06:12PM -0700, Gurucharan Shetty wrote: >> There is no 'kill -l' type functionality available on Windows. >> So instead of looking for the string 'ABRT', check for the exit >> code which when run on msys (unit test environment) is 9. > > We could unify this a bit: on Unix-like environments, the exit status > in the same circumstances should be 134 (SIGABRT + 128). So the > Windows versus Unix difference could just be the exit status we test > for. > > I don't know whether shell $variables are expanded in the "exit > status" parameter to AT_CHECK. Looks like you can expand $variables inside AT_CHECK's exit status. So I will use your recommendation.
> >> Also, after a call to abort(), on Windows, stderr does not get >> flushed to any file. So, do not look for it. > > I don't think that Unix-like systems flush stderr to a file after > abort(), either. I think that the reason that this works on Unix-like > systems is because stderr is line-buffered (not fully buffered) by > default. That is supposed to be the case on every compliant C > implementation (there are only three kinds of buffering: not buffered, > line buffered, fully buffered): You are correct. I made an incorrect reasoning. > > As initially opened, the standard error stream is not fully > buffered; the standard input and standard output streams are fully > buffered if and only if the stream can be determined not to refer > to an interactive device. > > But: if you add something to ovstest.c initialization that changes > stderr to line-buffered, does it make the need for this change go > away? Yes, this will work. I will have it as part of v2. > > Thanks, > > Ben. _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
