Error messages should go to STDERR, not STDOUT

[rpdelaney]

This would facilitate a wrapper script to pipe errors and so on. See http://google-styleguide.googlecode.com/svn/trunk/shell.xml#STDOUT_vs_STDERR

[Hasimir]

I'd need to double-check the spec, but I'm pretty sure UCI only deals with STDIN and STDOUT, not STDERR (possibly short-sighted, but not likely to be changed). In which case it may be necessary for errors to go to stdout as a result.

[rpdelaney]

That's a bummer. There are some possible hacks to work-around both standards, such as reading environment variables or parameters that instruct the script to implement non-standard behavior (in either direction). For example TB_USE_STDERR if non-zero would print errors to STDERR only, and TB_USE_ALL if non-zero would print errors to both STDOUT and STDERR. Default behaviors (if neither is set) would be to print all output to STDOUT.

But would it be necessary for the tb generation scripts to follow the UCI protocols in their output? I would think that would not be a concern in any way, unless there are GUI applications that invoke them to generate tablebases expecting a UCI interface...?

[Hasimir]

One of the reasons UCI has gained popularity in recent years is that it is a very simple and easy to understand interface. It's just text input and output which is, more or less, in plain English. It's not a massive protocol either , you should be able to read through the whole thing in under an hour.

[syzygy1]

Fixed.