Can you explain the rationale for the #if used with the same SILVERLIGHT setting?
It seems like it limits the code. On Sat, Nov 29, 2014 at 12:03 PM, Alexander Köplinger < alex.koeplin...@outlook.com> wrote: > Yeah, that was what I was doing. > > I created these PRs: > https://github.com/mono/referencesource/pull/1 > https://github.com/mono/mono/pull/1433 > > -- Alex > > ------------------------------ > From: mig...@xamarin.com > Date: Sat, 29 Nov 2014 11:23:01 -0500 > > Subject: Re: [Mono-dev] Porting System.Numerics from referencesource, > approaches for copying code? > To: alex.koeplin...@outlook.com > CC: mono-devel-list@lists.ximian.com > > I just realized it might not have been 100% clear what I think we should > do. > > We would in this case replace the references to our > mono/mcs/class/XX/*/*cs source files, with the > "../../external/referencesource/..." files, and then alter the original > referencesource file with the #if statements. > > We have done that already in a couple of places. > > (More to come, last night we got XML working, but needs some work) > > On Sat, Nov 29, 2014 at 11:21 AM, Miguel de Icaza <mig...@xamarin.com> > wrote: > > Thanks Alex. > > I do not think that we will have a strict rule for when to copy and when > to reference. > > There are scenarios where we will want to keep a lot of the code in one of > our files, and pull code from the referencesource. At that point, we will > have to determine whether the use of #ifdefs or partial classes is a better > long-term maintenance approach than copying the source. > > In this particular case, I feel that we can modify the reference source, > and add the #if MONO around it. > > The reason is that the changes are pretty straight forward, it is mostly > about disabling code, and not about picking some specific bits of code and > putting them on our tree. So an #if would be nice, since we can do a few > things (a) we can later find out what code is being disabled by the port > (looking at work we can do in the future to import new versions, or > contribute back to .NET) and (b) the changes are minimal right now. > > Miguel > > On Sat, Nov 29, 2014 at 10:59 AM, Alexander Köplinger < > alex.koeplin...@outlook.com> wrote: > > Here's the diff between my trimmed version and the original: > http://www.mergely.com/9E70e5Ls/ > > Hmm, I actually think that most of the removed methods don't matter (as > they aren't called anyway). I just wanted to make the file only contain the > code that is necessary. > The only "real" changes I had to make where accesses to internal fields in > NumberFormatInfo, I just replaced them with accesses to the public > properties that read them. > > Maybe it makes more sense to correct those accesses directly in the > referencesource, so we can reference it from there? > As a general guidance, when should we copy something to the Mono tree vs. > editing referencesource? And if we copy, should we keep unneeded methods or > remove them? > > -- Alex > > ------------------------------ > From: mig...@xamarin.com > Date: Sat, 29 Nov 2014 10:40:10 -0500 > Subject: Re: [Mono-dev] Porting System.Numerics from referencesource, > approaches for copying code? > To: alex.koeplin...@outlook.com > CC: mono-devel-list@lists.ximian.com > > > Hey Alex, > > In general, the approach works. > > I am curious as to what is that you had to remove, and why. Can you > describe the needed changes? (I saw the diff, but it is hard to tell what > happened with it). > > I thought that Numerics was pretty much standalone. > > On Sat, Nov 29, 2014 at 10:32 AM, Alexander Köplinger < > alex.koeplin...@outlook.com> wrote: > > Hey, I've been trying to port System.Numerics (BigInteger, Complex > classes) from referencesource to see and feel what the process looks like. > It mostly went fine, but there was one file (number.cs) I couldn't > reference directly from the submodule and had to copy to the assembly folder > in the Mono tree. > > I had to remove a bunch of code from that file to trim it down to only > what is needed for System.Numerics to compile (all the tests pass now as > well). > Question: is this a valid approach or is there something better? Note that > ifdefing the file wasn't feasible as it originally sits inside corlib and > has many other unrelated methods in it. > > Here's the commit with the WIP: > https://github.com/akoeplinger/mono/commit/d7c461155a975a35f76b79ebcc3fe33407bd5dce > > -- Alex > > _______________________________________________ > Mono-devel-list mailing list > Mono-devel-list@lists.ximian.com > http://lists.ximian.com/mailman/listinfo/mono-devel-list > > > > >
_______________________________________________ Mono-devel-list mailing list Mono-devel-list@lists.ximian.com http://lists.ximian.com/mailman/listinfo/mono-devel-list