From mboxrd@z Thu Jan 1 00:00:00 1970 From: Daniel Jacobowitz To: Andrew Cagney Cc: Fernando Nasser , gdb-patches@sources.redhat.com Subject: Re: [rfa/testsuite/mi] Fix transient failures in mi*-hack-cli.exp Date: Sun, 14 Oct 2001 16:47:00 -0000 Message-id: <20011014194805.A5331@nevyn.them.org> References: <20010928184755.A24096@nevyn.them.org> <3BB88356.8FD672CA@redhat.com> <3BB88CC7.6050405@cygnus.com> <3BCA222A.2050601@cygnus.com> X-SW-Source: 2001-10/msg00211.html On Sun, Oct 14, 2001 at 07:39:22PM -0400, Andrew Cagney wrote: > I wrote: > > >Um, I'd actually like to study this one in more detail > > Daniel wrote: > > >This one was a real nuisance to find... > > > >mi_gdb_test expects strings without the trailing newline. If there is a > >trailing newline, we'll get two prompts back. In some cases, depending on > >the delicacy of timing, this can cause a problem. The output looks like: > > > >47show architecture > > > >&"show architecture\n" > >~"The target architecture is assumed to be i386\n" > >47^done > >(gdb) > >&"\n" > >^done > >(gdb) > > > >Removing the unnecessary newlines masks the bug, as best as my tests can > >tell. OK to commit? > > (sorry for the delay). > > More to the point, it shouldn't send the blank line anyway > Yes, ok. I think 5.1 branch as well since a random pass/fail will > confuse people trying to report test results. OK, I'll do that on Monday. If no one objects (why would anyone object? :) I'm going to put my other recent testsuite fixes on the branch too; they fix timeout issues, and it's much more convenient to run the testsuite after they're applied. Is that OK? -- Daniel Jacobowitz Carnegie Mellon University MontaVista Software Debian GNU/Linux Developer