Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Thu, 5 Jul 2012 22:16:12 +0400
From: Solar Designer <>
Subject: Re: Should we add --make_check to --list=hidden-options?

On Thu, Jul 05, 2012 at 09:20:33AM -0500, jfoug wrote:
> make_check should be documented as a hidden option (or simply removed).  It
> has been hidden for a long time (did not work for a long time also, but was
> fixed a couple years back).  It really is deprecated by both -test=0  (the
> =0 part), and by the TS.  Test used to have no optional time value, thus
> would run full benchmark timings each time, no matter what.

I disagree that --make_check is deprecated - at least not in the core
tree.  It also was never broken in the core tree - or Owl package builds
would fail, which I'd notice before making a John release.  It is news
to me that it was broken in jumbo, but apparently no one cared if so.

"make check" and thus --make_check is used from Owl's john.spec to test
the various binaries that are built from there, with runtime fallback
suppressed - e.g., when testing the SSE2 binary on a compiling machine
that only has MMX, we get a "Sorry, SSE2 is required" in the build log,
yet the build continues (with the SSE2 binary remaining untested, but
packaged as if it passed the test).

--test=0 is not exactly the same.  It does not suppress the fallback,
and if the proper CPU is not detected and there's no fallback (or if it
fails), then --test=0's exit code would be 1, whereas --make_check's
would be 0.

As to adding it to --list=hidden-options, I don't mind, but I see little
need.  It is meant to be used by "make check" only.


Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.