# $NetBSD: varparse-errors.mk,v 1.9 2023/06/01 20:56:35 rillig Exp $ # Tests for parsing and evaluating all kinds of variable expressions. # # This is the basis for redesigning the error handling in Var_Parse and # Var_Subst, collecting typical and not so typical use cases. # # See also: # Var_Parse # Var_Subst PLAIN= plain value LITERAL_DOLLAR= To get a dollar, double $$ it. INDIRECT= An ${:Uindirect} value. REF_UNDEF= A reference to an ${UNDEF}undefined variable. ERR_UNCLOSED= An ${UNCLOSED variable expression. ERR_BAD_MOD= An ${:Uindirect:Z} expression with an unknown modifier. ERR_EVAL= An evaluation error ${:Uvalue:C,.,\3,}. # In a conditional, a variable expression that is not enclosed in quotes is # expanded using the mode VARE_UNDEFERR. # The variable itself must be defined. # It may refer to undefined variables though. .if ${REF_UNDEF} != "A reference to an undefined variable." . error .endif # As of 2020-12-01, errors in the variable name are silently ignored. # Since var.c 1.754 from 2020-12-20, unknown modifiers at parse time result # in an error message and a non-zero exit status. # expect+1: Unknown modifier "Z" VAR.${:U:Z}= unknown modifier in the variable name .if ${VAR.} != "unknown modifier in the variable name" . error .endif # As of 2020-12-01, errors in the variable name are silently ignored. # Since var.c 1.754 from 2020-12-20, unknown modifiers at parse time result # in an error message and a non-zero exit status. # expect+1: Unknown modifier "Z" VAR.${:U:Z}post= unknown modifier with text in the variable name .if ${VAR.post} != "unknown modifier with text in the variable name" . error .endif # Demonstrate an edge case in which the 'static' for 'errorReported' in # Var_Subst actually makes a difference, preventing "a plethora of messages". # Given that this is an edge case and the error message is wrong and thus # misleading anyway, that piece of code is probably not necessary. The wrong # condition was added in var.c 1.185 from 2014-05-19. # # To trigger this difference, the variable assignment must use the assignment # operator ':=' to make VarEvalMode_ShouldKeepUndef return true. There must # be 2 expressions that create a parse error, which in this case is ':OX'. # These expressions must be nested in some way. The below expressions are # minimal, that is, removing any part of it destroys the effect. # # Without the 'static', there would be one more message like this: # Undefined variable "${:U:OX" # #.MAKEFLAGS: -dv IND= ${:OX} # expect+2: Undefined variable "${:U:OX" # expect+1: Undefined variable "${:U:OX" _:= ${:U:OX:U${IND}} ${:U:OX:U${IND}} #.MAKEFLAGS: -d0 # Before var.c 1.032 from 2022-08-24, make complained about 'Unknown modifier' # or 'Bad modifier' when in fact the modifier was entirely correct, it was # just not delimited by either ':' or '}' but instead by '\0'. UNCLOSED:= ${:U:Q UNCLOSED:= ${:U:sh UNCLOSED:= ${:U:tA UNCLOSED:= ${:U:tsX UNCLOSED:= ${:U:ts UNCLOSED:= ${:U:ts\040 UNCLOSED:= ${:U:u UNCLOSED:= ${:U:H UNCLOSED:= ${:U:[1] UNCLOSED:= ${:U:hash UNCLOSED:= ${:U:range UNCLOSED:= ${:U:_ UNCLOSED:= ${:U:gmtime UNCLOSED:= ${:U:localtime