On Wed, 28 Oct 2009, Barry Smith wrote: > > On Oct 28, 2009, at 6:41 PM, Satish Balay wrote: > > > Barry, > > > > any objections to removing the 'x11' test from 'make test?' > > > > It fails for significant numer of installs [for users without working > > x11 etc..]. And the c/fortran/parallel tests should be the primary > > tests anyway.. > > The make rule for make tests only runs this test when ./configure has found > X11 and tested that it compiles and links properly. > > Thus it is not failing because of a failed X11 install. > > My understanding is that it ONLY fails because people don't have the proper > DISPLAY variable set. That is either > 1) DISPLAY does not exist or has garbage or > 2) DISPLAY points to a proper X server but because of permission issues or > firewalls the X client cannot connect to the X server. > > Does it fail for any other reason?
Or one is not running a X server on their desktop [and installs on a remote linux box] > > I hate to pander to complete idiots who cannot set up their X system > properly; but I guess you can move that test to a new test called > testx11 I guess some users don't need or use X11 directly [esp on Mac?]. And with PETSc - its secondary usage anyway [We install with X11 , but its primarily useful to users at debugging stage - when debugging matrix assembly with -mat_view_draw etc..]. Hence my proposal to remove it from 'make test'. Will move it to 'testx11' thanks, satish
