> On Feb 28, 2017, at 21:53, Rodney W. Grimes > <freebsd-...@pdx.rh.cn85.dnsmgr.net> wrote: > > -- Start of PGP signed section. > [ Charset UTF-8 unsupported, converting... ] >> >>> On Feb 28, 2017, at 21:38, Rodney W. Grimes >>> <freebsd-...@pdx.rh.cn85.dnsmgr.net> wrote: >>> >>> Would it be possible to not do this as 1000 commits? >> >> Yes, but I?m trying to do this in a way that allows me to backport all >> of the changes as easily as possible to ^/stable/10 (on the other side of >> the coin, I?ve run into dependent commits of other MFCs that required >> sweeping changes in order to make my changes possible). >> I do think your point is valid though and I can probably strike some >> middle ground with my changes. I think grouping it up by 2nd level directory >> (like I did with usr.bin/svn) makes the most sense. >> Thank you, > > Yes, that is what I was thinking. Note that I dont agree with this change, > but I have probably missed a discussion that is allowing it to move forward. > It makes the logfiles of make output now src tree possition dependent, > but I am probalby the only person in the universe that has ever run a > diff on the output of make world.
It was much the same way before, i.e. .CURDIR was a full path. My doing this achieves what you described, it helps dumb tools (like Jenkins) dedupe warnings though, and it also makes it easier to fork/build on different components in the source tree (I recently embarked on joining a forked copy of syslogd at $work, and using SRCTOP helps me achieve that, in the sense that I don’t need to duplicate Makefiles). > It may also be possible to split this into some phases to reduce some of > your issues, only do the {.CURDIR}/../../foo -> ${SRCTOP}/bar/foo in one pass > and > some of the other fixes in another pass. Yeah. That seems reasonable (I will split it up into larger chunks). Thanks, -Ngie
signature.asc
Description: Message signed with OpenPGP using GPGMail