Revision: arch--lord--1.0--patch-155
Archive: lord@regexps.com--2002
Creator: Tom Lord <lord@regexps.com>
Date: Wed Feb 27 04:01:54 PST 2002
Standard-date: 2002-02-27 12:01:54 GMT
Summary: `merge-points' bug fix
Keywords: 
New-files: {arch}/arch/arch--lord/arch--lord--1.0/lord@regexps.com--2002/patch-log/patch-155
Modified-files: ChangeLog
  ChangeLog.d/geisler@ece.nwu.edu--public/ChangeLog.geisler--1.0
  ChangeLog.d/geisler@ece.nwu.edu--src/ChangeLog.geisler--1.0
  ChangeLog.d/lord@regexps.com--2002/ChangeLog.lord--1.0
  ChangeLog.d/lord@regexps.com--2002/ChangeLog.lord-doc--1.0
  patch-logs/merge-points.sh
New-patches: lord@regexps.com--2002/arch--lord--1.0--patch-155


`merge-points' produces stdout output in a for loop.  When piped into
a process that only reads some of that output (e.g. `head -n 1') some
of the process invoked in that for loop get EPIPE or die on SIGPIPE,
apparently depending on the shell being used or the particular
implementation of the shell tool in question.  In such cases, some
implementations of the programs in question spew garbage on stderr
(EPIPE and SIGPIPE mean "stop sending output, nobody is reading it" --
it doesn't mean "something has gone disasterously wrong, please make
noise on stderr").  To work around those bugs, `merge-points' now
pipes its output through `pipe-friendly' which handles EPIPE with
greater grace.


