On Tue, Dec 03, 2019 at 10:30:35AM +0900, Amit Langote wrote:
> How about adding a function, say print_progress_to_stderr(const char
> *fmt,...), exposed to the front-end utilities and use it from
> everywhere? Needless to say that it will contain the check for whether
> stderr points to terminal or a file and print accordingly.
I have considered this point, but that does not seem worth the
complication as each tool has its own idea of the log output, its own
idea of the log output timing and its own idea of when it is necessary
to print the last newline when finishing to the output with '\r'.
> Considering Fabien's comment on this, we will have to check which
> other instances are printing information that is not very useful to
> look at line-by-line.
Thanks, applied the part for the initialization to HEAD. I got to
think about Fabien's point and it is true that for pgbench's
--progress not keeping things on the same line for a terminal has
advantages because the data printed is not cumulative: that's a
summary of the previous state printed which can be compared.
Note: the patch works on Windows, no problem.
--
Michael