Re: possible test setup problem (LSB-OS)

[Andrew Josey <ajosey@rdg.opengroup.org>]

Chris
Some more information--- btw this is the same problem as all the failures
that were commented as "ENAMETOOLONG problem" in previous failure
analyses.  There were lots of these in the VSX-PCTS tests, e.g. fopen 25,
freopen 26, remove 14, etc.

To answer the question on TEST_XPG, TEST_XPG is 0 if you configure POSIX96 mode
and is 500 if you configure UNIX98 mode, so on a system where PATH_MAX
excludes the null terminator, the ENAMETOOLONG tests will all fail in
UNIX98 mode but pass in POSIX96 mode.  I believe that UNIX98 mode is
the right mode for running VSX-PCTS and LSB-OS, because the LSB spec
says that the behaviour of the system calls and functions that they test
is specified by XSH5.  I.e. they all have an "[8]" next to them in the
tables in chapter 10. (Actually, that's not quite true - some, such as
mkfifo(), have a "[4]" - but I assume that's because the LSB specifies
additional requirements in those cases.)  In the particular case of the
PATH_MAX value, the LSB spec says that pathconf() behaves as specified
in XSH5, which means the value for PATH_MAX returned by pathconf(path,
_PC_PATH_MAX) must include the null terminator.

regards
Andrew