Opened 2 months ago
Closed 2 months ago
#70665 closed defect (fixed)
ngspice-lib fails to build: error: duplicate member 'component'
| Reported by: | barracuda156 | Owned by: | bpdegnan (Brian Degnan) |
|---|---|---|---|
| Priority: | Normal | Milestone: | |
| Component: | ports | Version: | 2.10.1 |
| Keywords: | Cc: | markemer (Mark Anderson) | |
| Port: | ngspice, ngspice-lib |
Description
Aside from two obvious breakages introduced by a recent update, which are trivially fixable (drop hardcoded arch flag, set required C/C++ standard), there is another issue:
Making all in dense
make[3]: Entering directory `/opt/local/var/macports/build/_opt_PPCSnowLeopardPorts_science_ngspice/ngspice-lib/work/ngspice-43/src/maths/dense'
CC dense.lo
CCLD libdense.la
make[3]: Leaving directory `/opt/local/var/macports/build/_opt_PPCSnowLeopardPorts_science_ngspice/ngspice-lib/work/ngspice-43/src/maths/dense'
Making all in KLU
make[3]: Entering directory `/opt/local/var/macports/build/_opt_PPCSnowLeopardPorts_science_ngspice/ngspice-lib/work/ngspice-43/src/maths/KLU'
CC libKLU_real_la-klu.lo
CC libKLU_real_la-klu_diagnostics.lo
CC libKLU_real_la-klu_extract.lo
CC libKLU_real_la-klu_free_numeric.lo
CC libKLU_real_la-klu_dump.lo
CC libKLU_real_la-klu_factor.lo
klu_factor.c: In function 'klu_factor':
klu_factor.c:397:10: warning: variable 'R' set but not used [-Wunused-but-set-variable]
397 | Int *R ;
| ^
CC libKLU_real_la-klu_kernel.lo
klu_diagnostics.c: In function 'klu_rgrowth':
klu_diagnostics.c:37:14: warning: variable 'Ui' set but not used [-Wunused-but-set-variable]
37 | Int *Q, *Ui, *Uip, *Ulen, *Pinv ;
| ^~
klu_diagnostics.c: In function 'klu_condest':
klu_diagnostics.c:182:9: warning: variable 'nblocks' set but not used [-Wunused-but-set-variable]
182 | Int nblocks, i, j, jmax, jnew, pend, n ;
| ^~~~~~~
klu_diagnostics.c:181:10: warning: variable 'R' set but not used [-Wunused-but-set-variable]
181 | Int *R ;
| ^
klu_diagnostics.c: In function 'klu_flops':
klu_diagnostics.c:425:21: warning: variable 'n' set but not used [-Wunused-but-set-variable]
425 | Int k, ulen, p, n, nk, block, nblocks, k1 ;
| ^
CC libKLU_real_la-klu_multiply.lo
CC libKLU_real_la-klu_refactor.lo
CC libKLU_real_la-klu_scale.lo
CC libKLU_real_la-klu_solve.lo
klu_kernel.c: In function 'prune':
klu_kernel.c:553:17: warning: variable 'Ux' set but not used [-Wunused-but-set-variable]
553 | Entry *Lx, *Ux ;
| ^~
CC libKLU_real_la-klu_sort.lo
CC libKLU_real_la-klu_tsolve.lo
klu_refactor.c: In function 'klu_refactor':
klu_refactor.c:34:36: warning: variable 'Offi' set but not used [-Wunused-but-set-variable]
34 | Int *P, *Q, *R, *Pnum, *Offp, *Offi, *Ui, *Li, *Pinv, *Lip, *Uip, *Llen,
| ^~~~
klu_refactor.c:34:29: warning: variable 'Offp' set but not used [-Wunused-but-set-variable]
34 | Int *P, *Q, *R, *Pnum, *Offp, *Offi, *Ui, *Li, *Pinv, *Lip, *Uip, *Llen,
| ^~~~
klu_refactor.c:34:10: warning: variable 'P' set but not used [-Wunused-but-set-variable]
34 | Int *P, *Q, *R, *Pnum, *Offp, *Offi, *Ui, *Li, *Pinv, *Lip, *Uip, *Llen,
| ^
CC libKLU_real_la-klu_utils.lo
CC libKLU_complex_la-klu.lo
klu_sort.c: In function 'klu_sort':
klu_sort.c:101:9: warning: variable 'n' set but not used [-Wunused-but-set-variable]
101 | Int n, nk, nz, block, nblocks, maxblock, k1 ;
| ^
CC libKLU_complex_la-klu_diagnostics.lo
CC libKLU_complex_la-klu_dump.lo
In file included from klu_internal.h:12,
from klu.c:61:
klu_version.h:341:14: error: duplicate member 'component'
341 | #define Imag component [1]
| ^~~~~~~~~
CC libKLU_complex_la-klu_extract.lo
make[3]: *** [libKLU_complex_la-klu.lo] Error 1
make[3]: *** Waiting for unfinished jobs....
In file included from klu_internal.h:12,
from klu_diagnostics.c:14:
klu_version.h:341:14: error: duplicate member 'component'
341 | #define Imag component [1]
| ^~~~~~~~~
klu_diagnostics.c: In function 'klu_z_rgrowth':
klu_diagnostics.c:37:14: warning: variable 'Ui' set but not used [-Wunused-but-set-variable]
37 | Int *Q, *Ui, *Uip, *Ulen, *Pinv ;
| ^~
klu_diagnostics.c: In function 'klu_z_condest':
klu_diagnostics.c:182:9: warning: variable 'nblocks' set but not used [-Wunused-but-set-variable]
182 | Int nblocks, i, j, jmax, jnew, pend, n ;
| ^~~~~~~
klu_diagnostics.c:181:10: warning: variable 'R' set but not used [-Wunused-but-set-variable]
181 | Int *R ;
| ^
klu_diagnostics.c: In function 'klu_z_flops':
klu_diagnostics.c:425:21: warning: variable 'n' set but not used [-Wunused-but-set-variable]
425 | Int k, ulen, p, n, nk, block, nblocks, k1 ;
| ^
make[3]: *** [libKLU_complex_la-klu_diagnostics.lo] Error 1
In file included from klu_internal.h:12,
from klu_dump.c:9:
klu_version.h:341:14: error: duplicate member 'component'
341 | #define Imag component [1]
| ^~~~~~~~~
make[3]: *** [libKLU_complex_la-klu_dump.lo] Error 1
In file included from klu_internal.h:12,
from klu_extract.c:12:
klu_version.h:341:14: error: duplicate member 'component'
341 | #define Imag component [1]
| ^~~~~~~~~
make[3]: *** [libKLU_complex_la-klu_extract.lo] Error 1
make[3]: Leaving directory `/opt/local/var/macports/build/_opt_PPCSnowLeopardPorts_science_ngspice/ngspice-lib/work/ngspice-43/src/maths/KLU'
make[2]: *** [all-recursive] Error 1
make[2]: Leaving directory `/opt/local/var/macports/build/_opt_PPCSnowLeopardPorts_science_ngspice/ngspice-lib/work/ngspice-43/src/maths'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/opt/local/var/macports/build/_opt_PPCSnowLeopardPorts_science_ngspice/ngspice-lib/work/ngspice-43/src'
make: *** [all-recursive] Error 1
Change History (11)
comment:1 follow-ups: 2 4 Changed 2 months ago by markemer (Mark Anderson)
comment:2 follow-up: 3 Changed 2 months ago by barracuda156
Replying to markemer:
We don't hardcode arch do we? Or is that in ngspice? I have zero way to test this, I haven't had a working PPC machine quite some time. If you have a way to fix it without breaking modern macOS, I'm more than happy to merge it.
We probably should set the required C++ standard no matter what.
The portfile added -m64, unconditionally. No idea what for, MacPorts normally handles flags correctly, and this port is not an exception.
Anyway, I have a fix for KLU issue.
- S. Do you know why my patch was removed? It is still needed, upstream fixed nothing ever since.
get_avail_mem_size.c:22:2: warning: #import is a deprecated GCC extension [-Wdeprecated] 22 | #import <mach/mach.h> | ^~~~~~ get_resident_set_size.c: In function 'getCurrentRSS': get_resident_set_size.c:114:33: error: storage size of 'info' isn't known 114 | struct mach_task_basic_info info; | ^~~~ get_avail_mem_size.c:23:2: warning: #import is a deprecated GCC extension [-Wdeprecated] 23 | #import <mach/mach_host.h> | ^~~~~~ get_resident_set_size.c:115:40: error: 'MACH_TASK_BASIC_INFO_COUNT' undeclared (first use in this function); did you mean 'TASK_BASIC_INFO_COUNT'? 115 | mach_msg_type_number_t infoCount = MACH_TASK_BASIC_INFO_COUNT; | ^~~~~~~~~~~~~~~~~~~~~~~~~~ | TASK_BASIC_INFO_COUNT get_resident_set_size.c:115:40: note: each undeclared identifier is reported only once for each function it appears in get_resident_set_size.c:116:40: error: 'MACH_TASK_BASIC_INFO' undeclared (first use in this function); did you mean 'TASK_BASIC_INFO'? 116 | if ( task_info( mach_task_self( ), MACH_TASK_BASIC_INFO, | ^~~~~~~~~~~~~~~~~~~~ | TASK_BASIC_INFO get_resident_set_size.c:116:5: warning: this 'if' clause does not guard... [-Wmisleading-indentation] 116 | if ( task_info( mach_task_self( ), MACH_TASK_BASIC_INFO, | ^~ get_resident_set_size.c:119:9: note: ...this statement, but the latter is misleadingly indented as if it were guarded by the 'if' 119 | return (unsigned long long) info.resident_size; | ^~~~~~ get_resident_set_size.c:114:33: warning: unused variable 'info' [-Wunused-variable] 114 | struct mach_task_basic_info info; | ^~~~ get_resident_set_size.c:140:1: warning: control reaches end of non-void function [-Wreturn-type] 140 | } | ^ make[4]: *** [get_resident_set_size.lo] Error 1
comment:3 follow-up: 5 Changed 2 months ago by markemer (Mark Anderson)
Replying to barracuda156:
Replying to markemer:
We don't hardcode arch do we? Or is that in ngspice? I have zero way to test this, I haven't had a working PPC machine quite some time. If you have a way to fix it without breaking modern macOS, I'm more than happy to merge it.
We probably should set the required C++ standard no matter what.
The portfile added
-m64, unconditionally. No idea what for, MacPorts normally handles flags correctly, and this port is not an exception.Anyway, I have a fix for KLU issue.
- S. Do you know why my patch was removed? It is still needed, upstream fixed nothing ever since.
get_avail_mem_size.c:22:2: warning: #import is a deprecated GCC extension [-Wdeprecated] 22 | #import <mach/mach.h> | ^~~~~~ get_resident_set_size.c: In function 'getCurrentRSS': get_resident_set_size.c:114:33: error: storage size of 'info' isn't known 114 | struct mach_task_basic_info info; | ^~~~ get_avail_mem_size.c:23:2: warning: #import is a deprecated GCC extension [-Wdeprecated] 23 | #import <mach/mach_host.h> | ^~~~~~ get_resident_set_size.c:115:40: error: 'MACH_TASK_BASIC_INFO_COUNT' undeclared (first use in this function); did you mean 'TASK_BASIC_INFO_COUNT'? 115 | mach_msg_type_number_t infoCount = MACH_TASK_BASIC_INFO_COUNT; | ^~~~~~~~~~~~~~~~~~~~~~~~~~ | TASK_BASIC_INFO_COUNT get_resident_set_size.c:115:40: note: each undeclared identifier is reported only once for each function it appears in get_resident_set_size.c:116:40: error: 'MACH_TASK_BASIC_INFO' undeclared (first use in this function); did you mean 'TASK_BASIC_INFO'? 116 | if ( task_info( mach_task_self( ), MACH_TASK_BASIC_INFO, | ^~~~~~~~~~~~~~~~~~~~ | TASK_BASIC_INFO get_resident_set_size.c:116:5: warning: this 'if' clause does not guard... [-Wmisleading-indentation] 116 | if ( task_info( mach_task_self( ), MACH_TASK_BASIC_INFO, | ^~ get_resident_set_size.c:119:9: note: ...this statement, but the latter is misleadingly indented as if it were guarded by the 'if' 119 | return (unsigned long long) info.resident_size; | ^~~~~~ get_resident_set_size.c:114:33: warning: unused variable 'info' [-Wunused-variable] 114 | struct mach_task_basic_info info; | ^~~~ get_resident_set_size.c:140:1: warning: control reaches end of non-void function [-Wreturn-type] 140 | } | ^ make[4]: *** [get_resident_set_size.lo] Error 1
We can add that patch back in, it built fine without it out it on arm and intel with v43. We can add it back in. I'll leave a comment for future me as to why it's in there.
comment:4 follow-ups: 6 7 Changed 2 months ago by barracuda156
Replying to markemer:
- S. Not that I expect anyone to test on powerpc when updating ports, that would be unreasonable, but closing a ticket about a failure was a bit premature.
comment:5 Changed 2 months ago by barracuda156
Replying to markemer:
We can add that patch back in, it built fine without it out it on arm and intel with v43. We can add it back in. I'll leave a comment for future me as to why it's in there.
Since there are several issues to address now, perhaps let me hopefully get the build succeeding, and I will tag you in a PR to review it.
(I mean, restoring a patch separately does not hurt, of course, it is just insufficient.)
comment:6 Changed 2 months ago by markemer (Mark Anderson)
Replying to barracuda156:
Replying to markemer:
- S. Not that I expect anyone to test on powerpc when updating ports, that would be unreasonable, but closing a ticket about a failure was a bit premature.
Yeah, I might have cast too wide a net when I closed those tickets. I'm just now noticing that one is pretty recent.
comment:7 Changed 2 months ago by markemer (Mark Anderson)
Replying to barracuda156:
Replying to markemer:
- S. Not that I expect anyone to test on powerpc when updating ports, that would be unreasonable, but closing a ticket about a failure was a bit premature.
Yeah, I don't want to break stuff if I don't have to - we need v43 for some reasonably modern models, but if it still builds with some tweaks, works for me.
comment:8 Changed 2 months ago by barracuda156
ngspice-lib is fixed, but ngspice still fails on the old bug:
CC ngspice-main.o
CC ngspice-conf.o
CC ngspice-ngspice.o
rm -f spinit spinit.tmp
srcdir=''; \
test -f ./spinit.in || srcdir=./; \
sed -e 's|@XSPICEINIT[@]||g' -e 's|@pkglibdir[@]|../lib/ngspice|g' ${srcdir}spinit.in >spinit.tmp
mv spinit.tmp spinit
{standard input}:1843:FATAL:incompatible feature used: directive .non_lazy_symbol_pointer (must specify "-dynamic" to be used)
make[2]: *** [ngspice-main.o] Error 1
make[2]: Leaving directory `/opt/local/var/macports/build/_opt_PPCSnowLeopardPorts_science_ngspice/ngspice/work/ngspice-43/src'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/opt/local/var/macports/build/_opt_PPCSnowLeopardPorts_science_ngspice/ngspice/work/ngspice-43/src'
make: *** [all-recursive] Error 1
comment:9 Changed 2 months ago by barracuda156
This is what fails, apparently:
spinitpath = ../lib/ngspice edit = sed \ -e 's|@XSPICEINIT[@]|$(XSPICEINIT)|g' \ -e 's|@pkglibdir[@]|$(spinitpath)|g' AM_CFLAGS = -static
Why does it even need a static linking?
UPD. I think we just need to let the linker do its job without forcing this flag, and it will work.
comment:11 Changed 2 months ago by Sergey Fedorov <barracuda@…>
| Resolution: | → fixed |
|---|---|
| Status: | assigned → closed |
Note: See
TracTickets for help on using
tickets.
We don't hardcode arch do we? Or is that in ngspice? I have zero way to test this, I haven't had a working PPC machine quite some time. If you have a way to fix it without breaking modern macOS, I'm more than happy to merge it.
We probably should set the required C++ standard no matter what.