Re: CVS commit: src/usr.bin/kdump
On 02.04.2020 23:35, Christos Zoulas wrote: > Ah debugging remnants. I'll remove it. Can you look at PR/55128? > OK! > static inline union savefpu * > fpu_lwp_area(struct lwp *l) > { > struct pcb *pcb = lwp_getpcb(l); > union savefpu *area = &pcb->pcb_savefpu; > > KASSERT((l->l_flag & LW_SYSTEM) == 0); > if (l == curlwp) { > fpu_save(); > } > KASSERT(!(l->l_md.md_flags & MDL_FPU_IN_CPU)); <- this will fire if > the debugger calls it with l != curlew and it uses cpu > > return area; > } > > >> On Apr 2, 2020, at 5:31 PM, Kamil Rytarowski wrote: >> >> On 02.04.2020 19:40, Christos Zoulas wrote: >>> +set -x >>> +AWK=gawk >>> + >> >> gawk? >
Re: CVS commit: src/usr.bin/kdump
Ah debugging remnants. I'll remove it. Can you look at PR/55128? static inline union savefpu * fpu_lwp_area(struct lwp *l) { struct pcb *pcb = lwp_getpcb(l); union savefpu *area = &pcb->pcb_savefpu; KASSERT((l->l_flag & LW_SYSTEM) == 0); if (l == curlwp) { fpu_save(); } KASSERT(!(l->l_md.md_flags & MDL_FPU_IN_CPU)); <- this will fire if the debugger calls it with l != curlew and it uses cpu return area; } > On Apr 2, 2020, at 5:31 PM, Kamil Rytarowski wrote: > > On 02.04.2020 19:40, Christos Zoulas wrote: >> +set -x >> +AWK=gawk >> + > > gawk? signature.asc Description: Message signed with OpenPGP
Re: CVS commit: src/usr.bin/kdump
On 02.04.2020 19:40, Christos Zoulas wrote: > +set -x > +AWK=gawk > + gawk?
re: CVS commit: src/usr.bin/kdump
> > On 2. Jun 2018, at 22:07, matthew green wrote: > > > > Module Name:src > > Committed By: mrg > > Date: Sat Jun 2 20:07:15 UTC 2018 > > > > Modified Files: > > src/usr.bin/kdump: mkioctls > > > > Log Message: > > just include for mkioctls. this works fine for me > > for several platforms and fixes the clang build. > > > > > > To generate a diff of this commit: > > cvs rdiff -u -r1.50 -r1.51 src/usr.bin/kdump/mkioctls > > This breaks the build: > > In file included from /build/devel/src/sys/net/if.h:99:0, > from kdump-ioctl.c:23: > /build/devel/src/sys/altq/if_altq.h:48:2: error: unknown type name 'kmutex_t' > kmutex_t *ifq_lock; > ^~~~ > > The real problem seems the sys include search path: > > #include <...> search starts here: > /build/devel/obj/dist.amd64/usr/X11R7/include/libdrm > /build/devel/obj/dist.amd64/usr/X11R7/include/pixman-1 > /build/devel/obj/dist.amd64/usr/X11R7/include > /build/devel/src/external/cddl/osnet/sys > /build/devel/src/external/cddl/osnet/dist/uts/common > /build/devel/src/usr.bin/ktrace > /build/devel/src/sys > /build/devel/obj/dist.amd64/usr/include/gcc-6 > /build/devel/obj/tools.amd64/lib/gcc/x86_64--netbsd/6.4.0/include-fixed > /build/devel/obj/dist.amd64/usr/include > > Most includes come from "cddl/osnet/sys/sys" which looks wrong. > > Suppose the search better starts with "-I${NETBSDSRCDIR}/sys" ... i remain unable to reproduce this on my own system. i guess either your idea, or, -I${DESTDIR}/usr/include before the cddl includes, and make them all last. FWIW, this patch works for me (in that, it doesn't break). please feel free to commit this or something similar that works for you, as i'm hesitant to commit something else here without confirming it helps the problem i don't see. thanks. .mrg. Index: Makefile.ioctl-c === RCS file: /cvsroot/src/usr.bin/kdump/Makefile.ioctl-c,v retrieving revision 1.35 diff -p -u -r1.35 Makefile.ioctl-c --- Makefile.ioctl-c28 May 2018 21:05:02 - 1.35 +++ Makefile.ioctl-c3 Jun 2018 11:43:19 - @@ -47,11 +47,6 @@ DPSRCS+= ${PROG}-ioctl.c CPPFLAGS+= -I${DESTDIR}/usr/X11R7/include/libdrm CPPFLAGS+= -I${DESTDIR}/usr/X11R7/include/pixman-1 CPPFLAGS+= -I${DESTDIR}/usr/X11R7/include -.if ${MKDTRACE} != "no" -CPPFLAGS+= -I${NETBSDSRCDIR}/external/cddl/osnet/sys -CPPFLAGS+= -I${NETBSDSRCDIR}/external/cddl/osnet/dist/uts/common -CWARNFLAGS+= -Wno-unknown-pragmas -.endif CPPFLAGS+= -D_DEV_IC_BT8XX_H_ CPPFLAGS+= -D_ALTQ_ALTQ_JOBS_H_# redefinition of inline # De-select one, dup ioctls @@ -65,4 +60,13 @@ CPPFLAGS+= -D__RADEON_DRM_H__ # Dup ioct CPPFLAGS+= -D__MACH64_DRM_H__ # Dup ioctls CPPFLAGS+= -D__MGA_DRM_H__ # Dup ioctls +# Make sure these are last. +.if ${MKDTRACE} != "no" +# We put /usr/include here explicitly to always prefer NetBSD headers. +CPPFLAGS+= -I${DESTDIR}/usr/include +CPPFLAGS+= -I${NETBSDSRCDIR}/external/cddl/osnet/sys +CPPFLAGS+= -I${NETBSDSRCDIR}/external/cddl/osnet/dist/uts/common +CWARNFLAGS+= -Wno-unknown-pragmas +.endif + ${DEPFILES}: .PRECIOUS
Re: CVS commit: src/usr.bin/kdump
> On 2. Jun 2018, at 22:07, matthew green wrote: > > Module Name: src > Committed By: mrg > Date: Sat Jun 2 20:07:15 UTC 2018 > > Modified Files: > src/usr.bin/kdump: mkioctls > > Log Message: > just include for mkioctls. this works fine for me > for several platforms and fixes the clang build. > > > To generate a diff of this commit: > cvs rdiff -u -r1.50 -r1.51 src/usr.bin/kdump/mkioctls This breaks the build: In file included from /build/devel/src/sys/net/if.h:99:0, from kdump-ioctl.c:23: /build/devel/src/sys/altq/if_altq.h:48:2: error: unknown type name 'kmutex_t' kmutex_t *ifq_lock; ^~~~ The real problem seems the sys include search path: #include <...> search starts here: /build/devel/obj/dist.amd64/usr/X11R7/include/libdrm /build/devel/obj/dist.amd64/usr/X11R7/include/pixman-1 /build/devel/obj/dist.amd64/usr/X11R7/include /build/devel/src/external/cddl/osnet/sys /build/devel/src/external/cddl/osnet/dist/uts/common /build/devel/src/usr.bin/ktrace /build/devel/src/sys /build/devel/obj/dist.amd64/usr/include/gcc-6 /build/devel/obj/tools.amd64/lib/gcc/x86_64--netbsd/6.4.0/include-fixed /build/devel/obj/dist.amd64/usr/include Most includes come from "cddl/osnet/sys/sys" which looks wrong. Suppose the search better starts with "-I${NETBSDSRCDIR}/sys" ... -- J. Hannken-Illjes - hann...@eis.cs.tu-bs.de - TU Braunschweig (Germa
Re: CVS commit: src/usr.bin/kdump
In article <20171106163453.gb28...@britannica.bec.de>, Joerg Sonnenberger wrote: >On Sun, Nov 05, 2017 at 12:44:29PM -0500, Christos Zoulas wrote: >> Module Name: src >> Committed By:christos >> Date:Sun Nov 5 17:44:29 UTC 2017 >> >> Modified Files: >> src/usr.bin/kdump: mkioctls >> >> Log Message: >> deal with the stdbool.h mess defining bool in and >> and then xf86Opt.h wanting to define a struct field called bool. > >I'd prefer to fix the latter. You mean change xf86Opt.h and rename 'bool' to 'xbool' or something else? That would probably require touching a lot of other X client code, and things will stop compiling out of the box for NetBSD. I think upstream should fix it first :-) christos
Re: CVS commit: src/usr.bin/kdump
On Sun, Nov 05, 2017 at 12:44:29PM -0500, Christos Zoulas wrote: > Module Name: src > Committed By: christos > Date: Sun Nov 5 17:44:29 UTC 2017 > > Modified Files: > src/usr.bin/kdump: mkioctls > > Log Message: > deal with the stdbool.h mess defining bool in and > and then xf86Opt.h wanting to define a struct field called bool. I'd prefer to fix the latter. Joerg