CVS [email protected]: linux
A message that is not in the archive (sent off-list or not yet received): 20000524012413.A5507@spock
On 23-May-00 Klaus Naumann wrote:
[offset.h problems with my recent CVS changes snipped]
Quoted 4 lines
Well, unfortunately this doesn't fix the problem. At least not for me. I get the same error I stated a mail ago -
offset.his generated after the use inmain.c.
Ok, trying to fix that I added a fastdep rule to arch/mips/tools/Makefile only to find myself in a catch 22 situation. offset.c includes asm/ptrace.h which, in turn, includes asm/offset.h, i.e. you have to have offset.h to create offset.h :-o
Without heavily messing around with several header files, which may have an impact on non-MIPS platforms as well, I see no easy solution for that. Adding an empty offset.h, on the other hand, and leaving the .cvsignore in place should at least partially do what I want. Any objections?
And now for something completly different. Why the hell did I mess around with that?
Well, if something in the kernel is changed which affects offset.h then generating a diff against a fresh CVS copy will contain those differences. That annoyed me because, personally, I find this disturbing when reviewing those diffs. Nobody really needs that anyway, at least that's what I thought, because offset.h is generated automatically.
The suggestion I made above would not solve that but at least avoids that those changes creep into the CVS.
Signature · 2 lines
Regards, Harald
Quoted 12 lines · Harald Koerfgen
Ok, trying to fix that I added a fastdep rule to
arch/mips/tools/Makefileonly to find myself in a catch 22 situation.offset.cincludesasm/ptrace.hwhich, in turn, includesasm/offset.h, i.e. you have to haveoffset.hto createoffset.h:-oWithout heavily messing around with several header files, which may have an impact on non-MIPS platforms as well, I see no easy solution for that. Adding an empty
offset.h, on the other hand, and leaving the .cvsignore in place should at least partially do what I want. Any objections?And now for something completly different. Why the hell did I mess around with that?
Quoted 8 lines
Well, if something in the kernel is changed which affects
offset.hthen generating a diff against a fresh CVS copy will contain those differences. That annoyed me because, personally, I find this disturbing when reviewing those diffs. Nobody really needs that anyway, at least that's what I thought, becauseoffset.his generated automatically.The suggestion I made above would not solve that but at least avoids that those changes creep into the CVS.
A good solution is important now that we have SMP. Toggling CONFIG_SMP affects offset.h and not for all variations of make invocations we actually have the guarantee that offset.h is being rebuilt before using it.
Ralf
On 24-May-00 Ralf Baechle wrote:
[offset.h]
Quoted 4 lines
A good solution is important now that we have SMP. Toggling
CONFIG_SMPaffectsoffset.hand not for all variations of make invocations we actually have the guarantee thatoffset.his being rebuilt before using it.
What about making offset.h to depend on $(TOPDIR)/.config?
This may cause unneccessary rebuilds of objects depending on offset.h after configuration changes but it forces offset.h to be rebuilt when CONFIG_SMP is toggled.
For example:
Makefile Thu May 25 19:08:48 2000 renamed from Makefile.old Thu May 25 19:05:08 2000 +1 -1
@@ -18,7 +18,7 @@ offset.h: offset.s sed -n '/^@@@/s///p' $^ >$@ -offset.s: offset.c +offset.s: offset.c $(TOPDIR)/.config clean: rm -f offset.[hs] $(TARGET).new
Signature · 2 lines
Regards, Harald
Quoted 9 lines · Harald Koerfgen
On 24-May-00 Ralf Baechle wrote:
[offset.h]Quoted 4 lines
A good solution is important now that we have SMP. Toggling
CONFIG_SMPaffectsoffset.hand not for all variations of make invocations we actually have the guarantee thatoffset.his being rebuilt before using it.
Quoted 5 lines
What about making
offset.hto depend on $(TOPDIR)/.config?This may cause unneccessary rebuilds of objects depending on
offset.hafter configuration changes but it forcesoffset.hto be rebuilt whenCONFIG_SMPis toggled.
Would you mind removing the include/asm-$(ARCH)/offset.h from the CVS repository, I get conflicts and stuff when I update. I don't see why we need to have a dummy file there.
Ulf
Quoted 10 lines · Ulf Carlsson
Quoted 10 lines · On 24-May-00 Ralf Baechle wrote:
[
offset.h]Quoted 4 lines
A good solution is important now that we have SMP. Toggling
CONFIG_SMPaffectsoffset.hand not for all variations of make invocations we actually have the guarantee thatoffset.his being rebuilt before using it.What about making
offset.hto depend on $(TOPDIR)/.config?
Quoted 7 lines
Quoted 3 lines
This may cause unneccessary rebuilds of objects depending on
offset.hafter configuration changes but it forcesoffset.hto be rebuilt whenCONFIG_SMPis toggled.Would you mind removing the include/asm-$(ARCH)/offset.h from the CVS repository, I get conflicts and stuff when I update. I don't see why we need to have a dummy file there.
Conceptually I don't like that either.
Ralf
Quoted 5 lines
Quoted 3 lines
Would you mind removing the include/asm-$(ARCH)/offset.h from the CVS repository, I get conflicts and stuff when I update. I don't see why we need to have a dummy file there.
Conceptually I don't like that either.
I think my solution that I submitted yesterday was pretty good. If add a rule that removes include/asm-$(ARCH)/offset.h when you make clean we will be just fine. You'll have to make clean when you toggle between SMP and UP, but I think we can live with that.
Ulf
Quoted 10 lines · Ulf Carlsson
Quoted 5 lines
Quoted 3 lines
Would you mind removing the include/asm-$(ARCH)/offset.h from the CVS repository, I get conflicts and stuff when I update. I don't see why we need to have a dummy file there.
Conceptually I don't like that either.
I think my solution that I submitted yesterday was pretty good. If add a rule that removes include/asm-$(ARCH)/offset.h when you make clean we will be just fine. You'll have to make clean when you toggle between SMP and UP, but I think we can live with that.
The recent change of __SMP__ to CONFIG_SMP almost everywhere in the kernel was intended to avoid just that.
Ralf