Hi Subhraman,
thanks for reporting this.
Two separate problems in libapl.cc:
1. release_value() had an invalid const_cast between Value and cValue
(two distinct classes, not the same type differently cv-qualified).
Since APL_value is Value*, "const APL_value val" means "Value* const"
(a const pointer to a non-const Value), not "pointer to const Value" —
val was never actually const, so the cast was wrong and unnecessary.
2. Once that compiles, libapl.cc hits ten more errors: several eval_*
calls still passed a raw Value* where the eval_AB/eval_ALB/eval_AXB/…
signatures now expect "const cValue &" (the cValue_R refactor from
*SVN r2025*). Fixed by dereferencing (*A instead of A.get()), matching
the idiom used elsewhere in the interpreter.
Also cleaned up the Function_PC/size_t signedness warning you quoted,
since this tree builds with -Werror.
Both apl and libapl.so now build cleanly here, including with -Werror.
Fixed in *SVN r2037*.
Best regards,
Jürgen
P.S. This reply was drafted with the help of Claude Code
(https://savannah.gnu.org/projects/apl).
On 7/12/26 06:47, Subhraman Sarkar wrote:
Hello,
Building apl succeeds, but cannot build libapl with the following error:
libapl.cc: In function ‘void release_value(APL_value, const char*)’:
libapl.cc:131:34: error: invalid ‘const_cast’ from type ‘const
APL_value’ {aka ‘Value* const’} to type ‘cValue*’
131 | Value * v = static_cast<Value *>(const_cast<cValue *>(val));
| ^~~~~~~~~~~~~~~~~~~~~~~~~
libapl.cc: In function ‘const Function* get_function_ucs(const
unsigned int*, const Function**, const Function**)’:
libapl.cc:574:44: warning: comparison of integer expressions of
different signedness: ‘Function_PC’ and
‘std::vector<Token>::size_type’ {aka ‘long unsigned int’}
[-Wsign-compare]
574 | for (Function_PC PC = Function_PC_0; PC < tos.size(); ++PC)
| ~~~^~~~~~~~~~~~
There are a lot more of those, on libapl.cc.
Build details:
Commit: 8b3cc77207021c472991ca476f61544da4bf96c8 (git, latest master)
OS: Devuan Excalibur
GCC: 14.2.0
Autoconf: autoconf (GNU Autoconf) 2.72
If any more details are needed, please ask.
