I have decided to largely modify the content of my post because the issue seems much clearer now than when I originally posted.
The task of using ccache for buildworld seems to have several strange problems associated with it. These are:
1. Suggested settings for make.conf do not work and ccache does not kick in when making buildworld. After I commented out the "if" statement ccache participated in the buildworld.
2. For amd64 systems, the i386-elf build seems to consistently break - posts about this issue exist for FreeBSD branch 8x also. Mostly suspected is a "compiler problem since it can't compile its own preprocessed output. A workaround for that is to set CCACHE_CPP2" (good example of the issue here).
3. Problem #2 did not clear up when I cleaned out the cache and updated devel/libtool. World builds fine without ccache and ommission or inclusion of -j option makes no difference.
4. I tried running ccache with CCACHE_CPP2 option for i386-elf only by placing in make.conf:
But this did not work. Probably once build starts, it does not re-consult make.conf for every depend so this setting was by-passed. Question: What is the command to start the build with CCACHE_CPP2 option for i386-elf only? How would I then integrate this into the buildworld? (buildworld must skip over this already compiled depend). My solution would be to edit i386-elf's Makefile and place the option there. Suggestions?
An alternatively acceptable solution here would be to build i386-elf without ccache (by-pass ccache for this module). Where & how could I specify that (still calls ccache if I just place in the i386-elf/Makefile
5. I think I should clear out all code already cached in ccache relating to i386-elf. Question: What would be the command for clearing that portion out only? Or is it risky and should I just start over completely?
UPDATE: I disabled ccache in make.conf and I ran make in /usr/src/lib/csu/i386-elf. It stopped with same error. I then deleted obj/src and started the build with disabled ccache to make sure - clean build. My conclusion therefore is that i386-elf with ccache breaks, not because of the conditions in i386-elf but because of code placed in ccache from previous modules being called for the build of i386-elf. The solution could be to not compile anything with ccache regarding 32bit src compatibility. Great - what happens when I try to cross-compile for an i386 machine (which I plan to do).
New Thought: Could a Custom Kernel cause this? Compat 32 is enabled (obviously).
The task of using ccache for buildworld seems to have several strange problems associated with it. These are:
1. Suggested settings for make.conf do not work and ccache does not kick in when making buildworld. After I commented out the "if" statement ccache participated in the buildworld.
Code:
[SIZE="3"].if (!empty(.CURDIR:M/usr/src*) || !empty(.CURDIR:M/usr/obj*)) && !defined(NOCCACHE)[/SIZE]
CC= /usr/local/libexec/ccache/world/cc
CXX= /usr/local/libexec/ccache/world/c++
.endif2. For amd64 systems, the i386-elf build seems to consistently break - posts about this issue exist for FreeBSD branch 8x also. Mostly suspected is a "compiler problem since it can't compile its own preprocessed output. A workaround for that is to set CCACHE_CPP2" (good example of the issue here).
3. Problem #2 did not clear up when I cleaned out the cache and updated devel/libtool. World builds fine without ccache and ommission or inclusion of -j option makes no difference.
4. I tried running ccache with CCACHE_CPP2 option for i386-elf only by placing in make.conf:
Code:
.if ${.CURDIR:M*/src/lib/csu/i386-elf}
CCACHE_CPP2= yes
.endifAn alternatively acceptable solution here would be to build i386-elf without ccache (by-pass ccache for this module). Where & how could I specify that (still calls ccache if I just place in the i386-elf/Makefile
Code:
NOCCACHE= yes5. I think I should clear out all code already cached in ccache relating to i386-elf. Question: What would be the command for clearing that portion out only? Or is it risky and should I just start over completely?
UPDATE: I disabled ccache in make.conf and I ran make in /usr/src/lib/csu/i386-elf. It stopped with same error. I then deleted obj/src and started the build with disabled ccache to make sure - clean build. My conclusion therefore is that i386-elf with ccache breaks, not because of the conditions in i386-elf but because of code placed in ccache from previous modules being called for the build of i386-elf. The solution could be to not compile anything with ccache regarding 32bit src compatibility. Great - what happens when I try to cross-compile for an i386 machine (which I plan to do).
New Thought: Could a Custom Kernel cause this? Compat 32 is enabled (obviously).