bug? handling <...>

[masatake]

I'm integrating pegof with the build system of Universal Ctags (https://github.com/universal-ctags/ctags/pull/4026).

During the integration, I found a bug of pegof.

[yamato@dev64]~/var/ctags-github% touch peg/varlink.peg  
[yamato@dev64]~/var/ctags-github% make V=1             
./misc/gen-repoinfo main/repoinfo.h
make  all-recursive
make[1]: Entering directory '/home/yamato/var/ctags-github'
Making all in gnulib
make[2]: Entering directory '/home/yamato/var/ctags-github/gnulib'
make  all-recursive
make[3]: Entering directory '/home/yamato/var/ctags-github/gnulib'
make[4]: Entering directory '/home/yamato/var/ctags-github/gnulib'
make[4]: Nothing to be done for 'all-am'.
make[4]: Leaving directory '/home/yamato/var/ctags-github/gnulib'
make[3]: Leaving directory '/home/yamato/var/ctags-github/gnulib'
make[2]: Leaving directory '/home/yamato/var/ctags-github/gnulib'
Making all in .
make[2]: Entering directory '/home/yamato/var/ctags-github'
./misc/gen-repoinfo main/repoinfo.h
../pegof/build/pegof -O all -o ./peg/varlink.pego -i "peg/varlink.peg"
./packcc -o ./peg/varlink ./peg/varlink.pego
gcc -DHAVE_CONFIG_H -I.  -I. -I. -I./main -I./dsl -I./peg -DHAVE_PACKCC  -DUSE_SYSTEM_STRNLEN -DHAVE_PEGOF -I./gnulib -I./gnulib -DHAVE_REPOINFO_H  -std=gnu99 -Wall   -I/usr/include/libxml2 -DWITH_GZFILEOP     -g -O2 -MT peg/libctags_a-varlink.o -MD -MP -MF peg/.deps/libctags_a-varlink.Tpo -c -o peg/libctags_a-varlink.o `test -f 'peg/varlink.c' || echo './'`peg/varlink.c
In file included from peg/varlink_pre.h:15,
                 from peg/varlink.c:28:
peg/varlink.c: In function ‘pcc_action_name_0’:
peg/varlink.c:1177:44: error: ‘_1’ undeclared (first use in this function); did you mean ‘_0’?
 1177 |     SET_SCOPE(auxil, makeVarlinkTag(auxil, _1, _1s));
      |                                            ^~
peg/peg_common.h:46:42: note: in definition of macro ‘SET_SCOPE’
   46 | #define SET_SCOPE(P,S) (BASE_SCOPE(P) = (S))
      |                                          ^
peg/varlink.c:1177:44: note: each undeclared identifier is reported only once for each function it appears in
 1177 |     SET_SCOPE(auxil, makeVarlinkTag(auxil, _1, _1s));
      |                                            ^~
peg/peg_common.h:46:42: note: in definition of macro ‘SET_SCOPE’
   46 | #define SET_SCOPE(P,S) (BASE_SCOPE(P) = (S))
      |                                          ^
peg/varlink.c:1177:48: error: ‘_1s’ undeclared (first use in this function); did you mean ‘_0s’?
 1177 |     SET_SCOPE(auxil, makeVarlinkTag(auxil, _1, _1s));
      |                                                ^~~
peg/peg_common.h:46:42: note: in definition of macro ‘SET_SCOPE’
   46 | #define SET_SCOPE(P,S) (BASE_SCOPE(P) = (S))
      |                                          ^
make[2]: *** [Makefile:6362: peg/libctags_a-varlink.o] Error 1
make[2]: Leaving directory '/home/yamato/var/ctags-github'
make[1]: *** [Makefile:6746: all-recursive] Error 1
make[1]: Leaving directory '/home/yamato/var/ctags-github'
make: *** [Makefile:2367: all] Error 2

I compared varlink.peg (the input for pegof) and varlink.pego (the output of pegof):

.peg

name
    <- < [A-Z][A-Za-z0-9]* > {
    if (PEEK_KIND (auxil) != KIND_GHOST_INDEX)
       SET_SCOPE(auxil, makeVarlinkTag(auxil, $1, $1s));
}

.pego

name <-
    [A-Z] [0-9A-Za-z]* { if (PEEK_KIND (auxil) != KIND_GHOST_INDEX)
       SET_SCOPE(auxil, makeVarlinkTag(auxil, $1, $1s)); }

pegof removes < and >. So $1 lost the sense. Should I turn unused-capture optimization off?

[dolik-rce]

Thanks for the detailed report @masatake.

Should I turn unused-capture optimization off?

Yes, you can try to identify the problematic optimization and turn it off, as a temporary workaround.

[masatake]

Thank you. All peg-based parsers are built well with -O all -X unused-capture. I will keep this one open.

[dolik-rce]

Should be fixed now, there was a stupid problem with the regexp that makes the check, so it didn't work correctly for multiline actions.