Re: [patch] minimal test case for sha-1 and md5

[Casey Marshall <csm at gnu dot org>]

>>>>> "Andreas" == Andreas Tobler <toa@pop.agri.ch> writes:

Andreas> Hi all,
Andreas> Ranjit Mathew wrote:

>>> 2004-10-12  Rutger Ovidius  <ovidr@users.sourceforge.net>
>>> 
>>> PR libgcj/17903:
>>> * testsuite/libjava.lang/md5test.java: New file.
>>> * testsuite/libjava.lang/md5test.out: Likewise.
>>> * testsuite/libjava.lang/shatest.java: Likewise.
>>> * testsuite/libjava.lang/shatest.out: Likewise.
>> Hi Andreas,
>> These tests fail for me because I run the testsuite
>> without installing the built GCJ:
>> -------------------------- 8< --------------------------
>> WARNING: could not properly read security provider files:
>> file:///home/ranmath/gcc/lib/security/libgcj.security
>> file:///home/ranmath/gcc/lib/security/classpath.security
>> Falling back to standard GNU security provider
>> -------------------------- 8< --------------------------

I seem to recall there being a mini flamewar about that message,
anyway: that java.security.Security prints messages to System.err if
it can't find any security properties file to load.

I think we are wrong to print that message, since nothing else in the
library does that, to my knowledge. Anyway, it's obnoxious, the
situation it describes just means a default was used, and it looks
more like debug cruft than anything else.

A nicer solution than following Sun's single-buttock'd way of doing
this would be to allow provider implementations to include this
information in Jar/.so files, like how JAXP does with
MEAT-INF/services.

Any votes for removing that message?

-- 
Casey Marshall || csm@gnu.org