Sender: rich AT phekda DOT freeserve DOT co DOT uk Message-ID: <3AA3FDD6.130C773F@phekda.freeserve.co.uk> Date: Mon, 05 Mar 2001 20:57:58 +0000 From: Richard Dawe X-Mailer: Mozilla 4.51 [en] (X11; I; Linux 2.2.17 i586) X-Accept-Language: de,fr MIME-Version: 1.0 To: DJGPP workers CC: Jim Meyering Subject: Re: Fetish.pm band-aid (Fileutils testsuite) References: Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Reply-To: djgpp-workers AT delorie DOT com Hello. Eli Zaretskii wrote: > I doubt that your guess is correct (modulo unknown bugs in the > library). Assuming that Perl calls `system' from libc, it should > handle quoted files just fine. This feature was heavily tested when > it was introduced (circa v2.01), and is working with no known problems > ever since. Yep, you're right Eli! > > "Ambiguous redirect" errors are generated. > > Both these messages ("Cannot redirect" and "Ambiguous redirect") are > from `system's code. Significantly, they both indicate that `system' > is using its internal emulator of COMMAND.COM, which doesn't support > "2>" style redirection. > > In other words, the test suite setup is not taking care of setting > $SHELL to point to the Bash executable. It should do that because > these commands will never work with anything but a Unixy shell. [snip] > It is quite common for Unix-born test suites to not bother setting > $SHELL, since the default shell is good enough. But for this to work > on non-Posix platforms, $SHELL _must_ be set. Well, the Makefiles have $SHELL set to /bin/sh, but that does not appear to be enough. When I tried with $SHELL set to /bin/sh.exe (which exists), I did not have the redirection problems. To configure Fileutils 4.0 and have the test suite avoid the redirection problems, I used the following: CONFIG_SHELL=$DJDIR/bin/bash.exe ./configure --disable-nls $CONFIG_SHELL needs to be set rather than $SHELL, because share/config.site overrides $SHELL. > If the failures with `dd' are not explained by stderr redirection, > please post the test source and results (diffs of actual against > expected output). It is impossible to reason about this without > knowing what does that test try to do. Sorry, I wasn't expecting anyone to suggest how to fix that. I should have been clearer. I will look at that next time I have my fileutils porting hat on. Thanks for the help, bye, Rich =] -- Richard Dawe http://www.bigfoot.com/~richdawe/ "The soul is the mirror of an indestructible universe." --- Gottfried W. Leibniz