Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Fri, 1 Jul 2011 12:02:41 -0500

Hi Ben,

On 2011-07-01 Benjamin Franksen wrote:
> 
> I am happy to announce the first release candidate of version 2.1.0 of the
> EPICS sequencer. You can download from here:

Testing against Base R3.14.12.1:

Building on solaris-sparc stops here:

/opt/SUNWspro/bin/CC -c  -D_POSIX_C_SOURCE=199506L  -D_XOPEN_SOURCE=500          
-DPVCA   -DUNIX  -DSOLARIS=10 -mt  -D__EXTENSIONS__   -O   +w           -
xcode=pic32  -I. -I../O.Common -I. -I.. -I../../../include/os/solaris -
I../../../include -I/usr/local/iocapps/R3.14.12.1/base/3-14-12-1-
asd1/include/os/solaris -I/usr/local/iocapps/R3.14.12.1/base/3-14-12-1-
asd1/include       ../pvCa.cc 
"../pvCa.cc", line 204: Error: The function "alloca" must have a prototype.
"../pvCa.cc", line 204: Error: Cannot use int to initialize char*.
"../pvCa.cc", line 272: Error: The function "alloca" must have a prototype.
"../pvCa.cc", line 272: Error: Cannot use int to initialize char*.
"../pvCa.cc", line 298: Error: The function "alloca" must have a prototype.
"../pvCa.cc", line 298: Error: Cannot use int to initialize char*.
"../pvCa.cc", line 477: Error: The function "alloca" must have a prototype.
"../pvCa.cc", line 477: Error: Cannot use int to initialize char*.

I can fix this with:

#if (__STDC_VERSION__ < 199901L) && !defined(__GNUC__)
#  include <alloca.h>
#endif

The conditional is necessary to avoid warnings from vxWorks, which doesn't 
have an alloca.h header (and I wonder if this might need a WIN32 equivalent as 
well, see below).

When it gets to compiling in src/snc I get lots of these warnings:

"../types.h", line 44: warning: typedef redeclared: uint

one of these:

"../analysis.c", line 1340: warning: integer overflow detected: op "<<"

and this:

"../gen_tables.c", line 185: warning: implicit function declaration: alloca

which results in this failure:

/opt/SUNWspro/bin/CC -o snc  -L/usr/local/iocapps/R3.14.12.1/base/3-14-12-1-
asd1/lib/solaris-sparc -R/usr/local/iocapps/R3.14.12.1/base/3-14-12-1-
asd1/lib/solaris-sparc         -L/usr/local/lib -mt  -z ignore -z combreloc -z 
lazyload  -R/usr/local/lib          lexer.o snl.o main.o expr.o var_types.o 
analysis.o gen_code.o gen_ss_code.o gen_tables.o sym_table.o    -lrecIoc -
lsoftDevIoc -lmiscIoc -lrsrvIoc -ldbtoolsIoc -lasIoc -ldbIoc -lregistryIoc -
ldbStaticIoc -lca -lCom -lCom   
Undefined			first referenced
 symbol  			    in file
alloca                              gen_tables.o
ld: fatal: Symbol referencing errors. No output written to snc

I can fix that with this 2-line #elif addition:

#ifdef _WIN32
#  include <malloc.h>
#elif (__STDC_VERSION__ < 199901L) && !defined(__GNUC__)
#  include <alloca.h>
#endif

I then get this failure when it gets to building in src/dev:

/opt/SUNWspro/bin/cc -c  -D_POSIX_C_SOURCE=199506L  -D_XOPEN_SOURCE=500            
-DUNIX  -DSOLARIS=10 -mt  -D__EXTENSIONS__   -xO4        -I../../seq       -
xcode=pic32  -I. -I../O.Common -I. -I.. -I../../../include/os/solaris -
I../../../include -I/usr/local/iocapps/R3.14.12.1/base/3-14-12-1-
asd1/include/os/solaris -I/usr/local/iocapps/R3.14.12.1/base/3-14-12-1-
asd1/include       ../devSequencer.c 
"../devSequencer.c", line 182: warning: implicit function declaration: alloca
/opt/SUNWspro/bin/CC -xar -o   libseqDev.a         devSequencer.o
/opt/SUNWspro/bin/CC -o libseqDev.so -z defs -G -h libseqDev.so -z text -
L/home/phoebus3/ANJ/epics/modules/soft/seq-2.1.0-rc1/lib/solaris-sparc -
L/usr/local/iocapps/R3.14.12.1/base/3-14-12-1-asd1/lib/solaris-sparc -
R/home/phoebus3/ANJ/epics/modules/soft/seq-2.1.0-rc1/lib/solaris-sparc -
R/usr/local/iocapps/R3.14.12.1/base/3-14-12-1-asd1/lib/solaris-sparc         -
L/usr/local/lib -mt  -z ignore -z combreloc -z lazyload  -R/usr/local/lib           
devSequencer.o   -lseq -lpv -lrecIoc -lsoftDevIoc -lmiscIoc -lrsrvIoc -
ldbtoolsIoc -lasIoc -ldbIoc -lregistryIoc -ldbStaticIoc -lca -lCom   -lposix4 
-lpthread    -lreadline -lcurses -lm -lCstd -lsocket -lnsl -lumem -lCrun -lc 
Undefined                       first referenced
 symbol                             in file
alloca                              devSequencer.o
ld: fatal: Symbol referencing errors. No output written to libseqDev.so
make[3]: *** [libseqDev.so] Error 1

but I can fix that just like in gen_tables.c

I also get these warnings from compiling bittypes.c:

/opt/SUNWspro/bin/cc -c  -D_POSIX_C_SOURCE=199506L  -D_XOPEN_SOURCE=500            
-DUNIX  -DSOLARIS=10 -mt  -D__EXTENSIONS__   -xO4                -I. -
I../O.Common -I. -I.. -I../../../include/os/solaris -I../../../include -
I/usr/local/iocapps/R3.14.12.1/base/3-14-12-1-asd1/include/os/solaris -
I/usr/local/iocapps/R3.14.12.1/base/3-14-12-1-asd1/include       bittypes.c 
"../bittypes.st", line 36: warning: initializer does not fit or is out of 
range: -1
"../bittypes.st", line 38: warning: initializer does not fit or is out of 
range: -1
"../bittypes.st", line 40: warning: initializer will be sign-extended: -1

After fixing all the build errors I ran the tests, which the first time 
resulted in an assert failure in cbLow:

norace.................ok 30/30                                              
A call to 'assert(pca)'
    by thread 'cbLow' failed in ../dbCa.c line 336.
EPICS Release EPICS R3.14.12.1-asd1 2011/04/26 21:43:15.
Local time is 2011-07-01 11:21:28.624182050 CDT
Please E-mail this message to the author or to [email protected]
Calling epicsThreadSuspendSelf()
Thread cbLow (466b8) suspended

Two subsequent re-runs both passed all tests without failing, although it is 
skipping 10 tests in pvGetIoc:

ok  2 - immediate completion:      expected=1==1=anon
ok  3 # SKIP no immediate completion
ok  4 - synchronous get:           expected=1==1=named

ok 14 - immediate completion:      expected=6==6=anon
ok 15 # SKIP no immediate completion
ok 16 - synchronous get:           expected=6==6=named

ok 26 - immediate completion:      expected=10==10=anon
ok 27 # SKIP no immediate completion
ok 28 - synchronous get:           expected=10==10=named

ok 38 - immediate completion:      expected=14==14=anon
ok 39 # SKIP no immediate completion
ok 40 - synchronous get:           expected=14==14=named

ok 50 - immediate completion:      expected=19==19=anon
ok 51 # SKIP no immediate completion
ok 52 - synchronous get:           expected=19==19=named

ok 62 - immediate completion:      expected=23==23=anon
ok 63 # SKIP no immediate completion
ok 64 - synchronous get:           expected=23==23=named

ok 74 - immediate completion:      expected=27==27=anon
ok 75 # SKIP no immediate completion
ok 76 - synchronous get:           expected=27==27=named

ok 86 - immediate completion:      expected=31==31=anon
ok 87 # SKIP no immediate completion
ok 88 - synchronous get:           expected=32==32=named

ok 97 - synchronous get:           expected=36==36=anon
ok 98 # SKIP no immediate completion
ok 99 - immediate completion:      expected=36==36=anon

ok 109 - synchronous get:           expected=40==40=anon
ok 110 # SKIP no immediate completion
ok 111 - immediate completion:      expected=40==40=anon

On solaris-x86 with the above changes I get the same results as solaris-sparc, 
plus one additional warning (which you can ignore):

/opt/SUNWspro/bin/cc -c  -D_POSIX_C_SOURCE=199506L  -D_XOPEN_SOURCE=500          
-D_X86_  -DUNIX  -DSOLARIS=10 -mt  -D__EXTENSIONS__   -xO4                -I. 
-I../O.Common -I. -I.. -I../../../include/os/solaris -I../../../include -
I/usr/local/iocapps/R3.14.12.1/base/3-14-12-1-asd1/include/os/solaris -
I/usr/local/iocapps/R3.14.12.1/base/3-14-12-1-asd1/include       lexer.c 
"./snl.h", line 73: warning: macro redefined: SS

On linux-x86_64 (RHEL 6) I get one build warning from snc (which I also saw on 
solaris BTW), but all of the runtests pass:

../syncq.st:16: warning: leaving out the queue size is deprecated, queue size 
defaults to 100 elements

On linux-x86 (which also cross-builds for rtems-mvme3100) I got the same 
warning for ../syncq.st, but no other issues and all the runtests passed.

The darwin-x86 system I use doesn't have re2c installed, I'll ask if it can be 
and try it when I can.

Thanks,

- Andrew
-- 
Optimization is the process of taking something that works and
replacing it with something that almost works, but costs less.
-- Roger Needham

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System