Re: Performance regression since make 3.81?

[David Rothlisberger]

On 15 April 2015 at 16:44, Paul Smith <address@hidden> wrote:
> You could generate a
> performance graph and see where make is spending its time during those
> 15 seconds.

oprofile says it's spending all its time in strcache_iscached. This is
called from verify_file_data_base, which is called from "die" in main.c.
This is why it was spending 10 seconds *after* make had already printed
"Nothing to be done".

"die" only calls verify_file_data_base if verify_flag is set, which is
only enabled if "-d" is set or if MAKE_MAINTAINER_MODE is set.

That explains why I was only seeing it in the versions of make I had
built myself from git, and not in make 3.81 from my distro.

So it isn't a regression! :-)

It might be worth mentioning the differences between a git build and
tarball build in README.git. (As far as I can tell, this, plus enabling
asserts, are the only differences.)

Dave.