Arch Linux User Repository

@feilen - perhaps a typeo on my part in the comment. In the PKGUBILD, I use the following: KCFLAGS KCPPFLAGS I believe these are correct. Is this not the case?

@graysky That is still not what I'm talking about. Here, maybe this will make it more clear: >KCCPFLAGS >KCPPFLAGS One of those two is the correct variable to use, but in one post you said KCCPFLAGS, while in the PKGBUILD it was KCPPFLAGS.

@graysky, hehe, seems like makepkg.conf flags don't affect the size image of my kernel image after all. My stupid ego is having problems dealing with this (or at least was having). Well, moving on. The '-O2' flag messes up "CONFIG_CC_OPTIMIZE_FOR_SIZE=y" (-Os) (funnily even resulted in a 96 bytes bigger kernel for me - 300 KiB smaller after removing '-O2') so at least that should be removed as it's default anyway. In addition as told in FS#11864, "-fstack-protector" has been enabled already in the kernel for a long time: https://bugs.archlinux.org/task/18864. I even tested both of these just to make sure. The KC(PP)FLAGS aren't needed in the package() function either since there's no compiling done there anyways. That line only installs the already built modules into the correct place: $\O/> make help|grep modules_i modules_install - Install all modules to INSTALL_MOD_PATH (default: /) Anyway, thanks for telling about these KC(PP)FLAGS. I honestly always thought that makepkg.conf flags would always be used by default.

@Det and gravsyk: Thank you very much for your help! Det was right. My makepkg.conf is exactly like the one from you (gravsky), so it was upgraded the right way. Thank you very much for your help!

@Det - The main reason why I bumped it to -4 is that I need to rebuild all of the repo packages, and to make that a meaningful exercise, the pkgver needs to increase or else pacman won't pick up that something changed and trigger an update. I want to keep the AUR and repo packages "in sync" so bumping to force a rebuild is required. To comment on your 2nd point about the long string for KCFLAGS/KCPPFLAGS: Since the KCFLAGS and KCPPFLAGS are as I understand it defined by the Makefile (or similar) included with the linux source. As I understand it, unless a user implicatively defines them at the make step, the generic defaults are used. Please correct me if I'm mistaken. Also, I do not believe that march=native provides the ARCH defaults of "-fstack-protector --param=ssp-buffer-size=4 -D_FORTIFY_SOURCE=2" but also please correct me if I'm mistaken. @Loth - My bad for pointing you to the pre-files. See an example of /etc/makepkg.conf here (adjust for your $arch): http://pastebin.com/YGqujRL2

@Lothium, it seems like you just fetched makepkg.conf from the [pacman] sources, through ABS or from here: http://projects.archlinux.org/svntogit/packages.git/tree/pacman/trunk/makepkg.conf. Either way, a simple "fix" would be to just remove your makepkg.conf and let pacman install the right one with: # pacman -Syy pacman E: wait, no, forget about that. It was just me not paying attention. The "@CARCH"and "@CHOST" stuff are just there to be changed in build time based on whether you are on i686 or x86_64. Nothing to worry there and definitely don't just copy+paste that file to your makepkg.conf But if you don't have a makepkg.conf.pacnew file in /etc - that's perfectly fine. The confs are prefixed like that only when the previous file had changed since the previous update (meaning _you_ changed it). So when you haven't changed the file in any way it can simply be replaced with the new one.

@gravsky: Thanks for your answer! I looked at the link you posted and checked the new makepkg.conf for a difference to mine. The only thing I found was that in my makepkg.con is CARCH="x86_64" CHOST="x86_64-unknown-linux-gnu" and in the new one there is CARCH="@CARCH@" CHOST="@CHOST@" Do I have to change mine to the new one? Best regards and thanks for your help!

@graysky, seems like you didn't understand the rest of my post either. What I meant was that there's no need to bump the pkgrel like that when the users should rebuild their package. It is a general rule in the AUR that when something breaks and requires a rebuild it's in the user's responsibility to rebuild the package. There's also the thing that [testing] users might've already rebuilt their package so when the maintainer bumps the pkgrel it's not really... nice. It's of course your choice. Your package, your rules but I'd just appreciate it, if you just made a comment about something like that. In this case the rebuild wasn't even 'required'. And the other thing was that instead of: "-march=native -O2 -pipe -fstack-protector --param=ssp-buffer-size=4 -D_FORTIFY_SOURCE=2" it would be perfecty fine (and more accurate) to use just: "-march=native" I explained this in my previous post.