Re: svn commit: r921181 - in /subversion/trunk/subversion/bindings/javahl/native: CopySources.h JNIThreadData.h RevisionRange.h RevpropTable.h
From: Hyrum K. Wright <hyrum_wright_at_mail.utexas.edu>
Date: Tue, 9 Mar 2010 17:05:58 -0600
On Mar 9, 2010, at 4:46 PM, Blair Zajac wrote:
> On 03/09/2010 02:41 PM, hwright_at_apache.org wrote:
I try to stay as far away from Pooo.h has possible. :)
In all seriousness, when working in the Java bindings, I spend a fair amount of time tracking down warnings and making sure they aren't errors. g++ can be verbose as it is; I'm just trying to make my life a little easier by eliminating a number of superfluous warnings. Since I seem to be the only person hacking JavaHL these days, I think this little bit of selfish indulgence is justified.
> I suggest just reverse merging this change, as while it's not the cost that matters (which is low as you state), but just being correct about it.
This point I understand. The problem happens when the class isn't declared in the proper order in the header sequence, and we end up with something that expands to:
The last line creates a warning, but only occasionally, depending on where the declaration falls in relation to various other includes (which are often in other files). The available options are:
I chose (2), which gives the benefits desired with the least amount of work. The problem really goes much farther than SVN::Pool: we include a number of our header files in other header files. To properly solve the problem requires (4); reverting r921181 doesn't really fix anything, it's just shuffling deck chairs.
If anybody wants to tackle (4), please feel free.
This is an archived mail posted to the Subversion Dev mailing list.