commits 7f452687 and cedf154ea6

Greg Banks gnb at fastmail.fm
Tue Feb 19 19:14:14 EST 2013



On Wed, Feb 20, 2013, at 09:13 AM, Дилян Палаузов wrote:
> Hallo Greg,
> 
> I checked very fast your commit from yesterday and ...

Thanks Dilyan, I greatly appreciate all reviews :)  I've cc'd the
development list to help show that review is good and should be public,
hope you don't mind.

> commit cedf154ea6dc3e0350023b8cf74e4b1ac8bb4484
> Author: Greg Banks <gnb at fastmail.fm>
> Date:   Mon Jul 2 20:59:32 2012 +1000
> 
>      configure checks for MySQL sensibly
> 
> 
> ... removes the variable $use_mysql from configure.ac, but $use_mysql is 
> used on the third last line in configure.ac to show if mysql was found;

Ah, this is the danger of merging in months-old commits.  That code at
the end of configure.ac did not exist when I was first writing my
commit.

So I think the end summary is really kind of pointless.  For example,
under "External dependencies" only the PCRE library is listed, but Cyrus
has *heaps* of external libraries it can use in various situations to
provide extra features, like krb5, openssl, uuid, gssapi and even zephyr
(whatever that is).  

But I don't see any good reason to break it if other people find it
useful, so please see commit "ea0c572 Fix configure summary for MySQL"

> 
> commit 7f452687a563aff45da070647a6ad32936f13785
> Author: Greg Banks <gnb at fastmail.fm>
> Date:   Mon Jul 9 15:05:46 2012 +1000
> 
>      util: fix subtle bug in buf_replace_*()
> 
>      when given a buf which has been initialised with buf_init_ro_cstr(),
>      buf_replace_*() will ensure that the target buf is a C string (for
>      searching) but not that it's writable (for replacing) which are
>      sometimes but not always the same thing.
> 
> 
> 
> ... adds #include "assert.c" to lib/util.c, but util.c already contains 
> #include <assert.c> by the time the commit was pushed at
> git.cyrusimap.org .
> 

Another subtle merge issue, thanks for finding that.  I've removed the
<assert.h>, as we want to be obviously and explicitly including
lib/assert.h not /usr/include/assert.h, in a way which will break at
compile time should someone later break the -I compile options. I also
went around and fixed some other #includes of a similar nature.  See
commit "cad2b6e Normalise use of #include "assert.h""


-- 
Greg.


More information about the Cyrus-devel mailing list