Hi Milan and All, >> > 1) Makefile.master > [..] > a) all those variables are used in gate somewhere
They were not in the version before my change, and every structured definition should try to be complete, so I don't see why Makefile.master should be required to only define variables (macros, actually) which are referenced somewhere. > b) if they are in separate change, because many of them have no impact > on your PostgreSQL I have opened another bug for the cleanup and I have added a comment in the webrev that this change is regarding the cleanup bug. If the person committing the change thinks that it should be a committed separately, then that is fine for me. > I understand what you are trying to do. SFW is not in the best shape, > its infrastructure is improving just these days (finally), but I would > prefer to make your changes more tied to PSQL build process. Maybe you > can discuss this particular change with Norm Jacobs who is trying to > clean up SFW I think. Sorry, I am not interested. I don't want this discussion. I wanted to make a little contribution to improve things and if that is not what you want then fine. In my next webrev, I will keep the changes which I think are absolutely necessary and not bother about the rest. On the other hand, I really don't understand what the problem is here. (Makefile.master) >> I can do without the copyright there Like I said, if that's a problem, I can do without. >> > 2) Targetdirs >> > >> > Targetdirs is neverending story. Nobody knows if all directory have to be >> there or not. By our discussion and what I remember from Mike Sulivan email, >> There should be only directories which are not created automatically. >> Personally >> I think there should be only directories which are not created by sfw gate >> like >> /var /usr and so on. >> > >> > Please, remove all modifications here. They are useless. >> >> Then please remove *everything* which is useless. I have not introduced this >> file and as long as it's there (what I have found when I started working on >> this >> was already more than what is not being created automatically), I adjust to >> what >> exists. >> > > Then open new CR requesting removal of unneeded lines in Targetdirs. I don't get this. Why are you asking ME to do this when all I wanted to do is to confirm with what was already there? >> > 3) METADATA >> > >> > Please, remove # comments first three lines. >> >> Why would you want to remove an extremely helpful comment which directs >> every >> developer to essential documentation regarding cryptic field names (like >> OSR, >> for instance)? >> > > Because it is not correct place where to document such thing. It can be > in some global README in gate or on opensolaris.org So are comments in source code not the correct place to document things? > Why should these 3 lines be in every METADATA file in gate? It is > wasting of resources. Please. You must be joking. >> > 6) libpqxx/Makefile.sfw >> > >> > libpqxx is not possible to make separately. It is annoying. We made >> conclusion that Makefile.sfw should be "makeable". Probably best way how to >> do >> it is to include some file which will contains path settings. >> >> OK, I can care about this later. Please get this thing done at least and I >> will >> be happy to make further contributions with further improvements. >> >> I Think I have already cleaned up libpqxx/Makefile.sfw significantly. >> > > In such case, could you document it in that file, please? Just like you > did with FIXME upper. OK. >> > 7) psqlrc issue >> > >> > After long and hot discussion we made a decision to do not accept this >> change >> for 8.3 at all. The reason is that it change behavior and there is no easy >> and >> safe way how to implemented and does not break current installations. >> >> See the public discussion. I don't agree with your argument, but you'll have >> the >> last word anyway. >> > > The last word will have CRT advocate :-) I must say the original > implementation was wrong and we spent many minutes to discuss clean way, > but the key reason for "not do it now" is in something which has some > interface impact (even if not documented now) and should be ARCed. And I > would prefer to not prolong this PSQL makefile cleanup for not so > important part. I got that point, yes. But still I don't like the idea not to fix things just because it's too much work. This is what generates even more work in the general case, but in this particular case it's really not worth it. Nils
