Hi Yes indeed I sent two versions intentionally the first with T. Fournier proposal the another one what I had in mind so at least any of them can be chosed freely but deeply that does not change much the things it is just "formal" ...
1/ Apart of popcount conflict, nothing particular that work pretty fine otherwise. 2/ I get your point, if Haproxy folks prefer just renaming popcount internal function to haproxy_popcount or whatever, it is ok for me then, I am quite open to other options :-) Kind regards. On 30 June 2015 at 09:42, Willy Tarreau <[email protected]> wrote: > Hello David, > > On Sun, Jun 28, 2015 at 01:33:49PM +0100, David CARLIER wrote: > > Here there are new patchiest with both versions whichever suits the best > > can be picked up. > > Thanks for your patches. Note that you sent two different versions of > patch 2. > > I have a few questions here : > > 1) I used to run haproxy on netbsd a long time ago, my front reverse > proxy was running on it before switching to openbsd. For a long time, > the openbsd target was perfectly usable to build for netbsd. Have you > found something which breaks when doing so nowadays and which requires > a dedicated target ? I'm not opposed to adding new ones even if that's > just to make things clearer, I'm trying to find the motivations to guess > if this is something that needs to be backported to other branches. > > 2) what's the purpose of changing the popcount implementation for the one > provided by the system ? As indicated in the comment, it's a "simple" > implementation, meaning that it doesn't try to be fast, efficient nor > anything, it's just simple for basic usage. It's only used when parsing > the config, so that's why I'm having difficulties understanding the > purpose of replacing it thus fragmenting the code base between > platforms. > > Thanks, > Willy > >
