1# $NetBSD: directive-for-escape.mk,v 1.16 2022/06/12 16:09:21 rillig Exp $ 2# 3# Test escaping of special characters in the iteration values of a .for loop. 4# These values get expanded later using the :U variable modifier, and this 5# escaping and unescaping must pass all characters and strings effectively 6# unmodified. 7 8.MAKEFLAGS: -df 9 10# Even though the .for loops take quotes into account when splitting the 11# string into words, the quotes don't need to be balanced, as of 2020-12-31. 12# This could be considered a bug. 13ASCII= !"\#$$%&'()*+,-./0-9:;<=>?@A-Z[\]_^a-z{|}~ 14 15# XXX: As of 2020-12-31, the '#' is not preserved in the expanded body of 16# the loop. Not only would it need the escaping for the variable modifier 17# ':U' but also the escaping for the line-end comment. 18.for chars in ${ASCII} 19. info ${chars} 20.endfor 21 22# As of 2020-12-31, using 2 backslashes before be '#' would treat the '#' 23# as comment character. Using 3 backslashes doesn't help either since 24# then the situation is essentially the same as with 1 backslash. 25# This means that a '#' sign cannot be passed in the value of a .for loop 26# at all. 27ASCII.2020-12-31= !"\\\#$$%&'()*+,-./0-9:;<=>?@A-Z[\]_^a-z{|}~ 28.for chars in ${ASCII.2020-12-31} 29. info ${chars} 30.endfor 31 32# Cover the code in ExprLen. 33# 34# XXX: It is unexpected that the variable V gets expanded in the loop body. 35# The double '$$' should intuitively prevent exactly this. Probably nobody 36# was adventurous enough to use literal dollar signs in the values of a .for 37# loop, allowing this edge case to go unnoticed for years. 38# 39# See for.c, function ExprLen. 40V= value 41VALUES= $$ $${V} $${V:=-with-modifier} $$(V) $$(V:=-with-modifier) 42.for i in ${VALUES} 43. info $i 44.endfor 45 46 47# Try to cover the code for nested '{}' in ExprLen, without success. 48# 49# The value of the variable VALUES is not meant to be a variable expression. 50# Instead, it is meant to represent literal text, the only escaping mechanism 51# being that each '$' is written as '$$'. 52VALUES= $${UNDEF:U\$$\$$ {{}} end} 53# 54# The .for loop splits ${VALUES} into 3 words, at the space characters, since 55# the '$$' is an ordinary character and the spaces are not escaped. 56# Word 1 is '${UNDEF:U\$\$' 57# Word 2 is '{{}}' 58# Word 3 is 'end}' 59# 60# Each of these words is now inserted in the body of the .for loop. 61.for i in ${VALUES} 62# $i 63.endfor 64# 65# When these words are injected into the body of the .for loop, each inside a 66# '${:U...}' expression, the result is: 67# 68# expect: For: loop body: 69# expect: # ${:U\${UNDEF\:U\\$\\$} 70# expect: For: loop body: 71# expect: # ${:U{{\}\}} 72# expect: For: loop body: 73# expect: # ${:Uend\}} 74# expect: For: end for 1 75# 76# The first of these expressions is the most interesting one, due to its many 77# special characters. This expression is properly balanced: 78# 79# Text Meaning Explanation 80# \$ $ escaped 81# { { ordinary text 82# UNDEF UNDEF ordinary text 83# \: : escaped 84# U U ordinary text 85# \\ \ escaped 86# $\ (expr) an expression, the variable name is '\' 87# \$ $ escaped 88# 89# To make the expression '$\' visible, define it to an actual word: 90${:U\\}= backslash 91.for i in ${VALUES} 92. info $i 93.endfor 94# 95# expect-3: ${UNDEF:U\backslash$ 96# expect-4: {{}} 97# expect-5: end} 98# 99# FIXME: There was no expression '$\' in the original text of the variable 100# 'VALUES', that's a surprise in the parser. 101 102 103# Second try to cover the code for nested '{}' in ExprLen. 104# 105# XXX: It is not the job of ExprLen to parse an expression, it is naive to 106# expect ExprLen to get all the details right in just a few lines of code. 107# Each variable modifier has its own inconsistent way of parsing nested 108# variable expressions, braces and parentheses. (Compare ':M', ':S', and 109# ':D' for details.) The only sensible thing to do is therefore to let 110# Var_Parse do all the parsing work. 111VALUES= begin<$${UNDEF:Ufallback:N{{{}}}}>end 112.for i in ${VALUES} 113. info $i 114.endfor 115 116# A single trailing dollar doesn't happen in practice. 117# The dollar sign is correctly passed through to the body of the .for loop. 118# There, it is expanded by the .info directive, but even there a trailing 119# dollar sign is kept as-is. 120.for i in ${:U\$} 121. info ${i} 122.endfor 123 124# As of 2020-12-31, the name of the iteration variable can even contain 125# colons, which then affects variable expressions having this exact modifier. 126# This is clearly an unintended side effect of the implementation. 127NUMBERS= one two three 128.for NUMBERS:M*e in replaced 129. info ${NUMBERS} ${NUMBERS:M*e} 130.endfor 131 132# As of 2020-12-31, the name of the iteration variable can contain braces, 133# which gets even more surprising than colons, since it allows to replace 134# sequences of variable expressions. There is no practical use case for 135# this, though. 136BASENAME= one 137EXT= .c 138.for BASENAME}${EXT in replaced 139. info ${BASENAME}${EXT} 140.endfor 141 142# Demonstrate the various ways to refer to the iteration variable. 143i= outer 144i2= two 145i,= comma 146.for i in inner 147. info . $$i: $i 148. info . $${i}: ${i} 149. info . $${i:M*}: ${i:M*} 150. info . $$(i): $(i) 151. info . $$(i:M*): $(i:M*) 152. info . $${i$${:U}}: ${i${:U}} 153. info . $${i\}}: ${i\}} # XXX: unclear why ForLoop_SubstVarLong needs this 154. info . $${i2}: ${i2} 155. info . $${i,}: ${i,} 156. info . adjacent: $i${i}${i:M*}$i 157.endfor 158 159# The variable name can be a single '$' since there is no check on valid 160# variable names. ForLoop_SubstVarShort skips "stupid" variable names though, 161# but ForLoop_SubstVarLong naively parses the body of the loop, substituting 162# each '${$}' with an actual 'dollar'. 163.for $ in dollar 164. info eight $$$$$$$$ and no cents. 165. info eight ${$}${$}${$}${$} and no cents. 166.endfor 167# Outside a .for loop, '${$}' is interpreted differently. The outer '$' starts 168# a variable expression. The inner '$' is followed by a '}' and is thus a 169# silent syntax error, the '$' is skipped. The variable name is thus '', and 170# since since there is never a variable named '', the whole expression '${$}' 171# evaluates to an empty string. 172closing-brace= } # guard against an 173${closing-brace}= <closing-brace> # alternative interpretation 174.info eight ${$}${$}${$}${$} and no cents. 175 176# What happens if the values from the .for loop contain a literal newline? 177# Before for.c 1.144 from 2021-06-25, the newline was passed verbatim to the 178# body of the .for loop, where it was then interpreted as a literal newline, 179# leading to syntax errors such as "Unclosed variable expression" in the upper 180# line and "Invalid line type" in the lower line. 181.for i in "${.newline}" 182. info short: $i 183. info long: ${i} 184.endfor 185 186# No error since the newline character is not actually used. 187.for i in "${.newline}" 188.endfor 189 190# Between for.c 1.161 from 2022-01-08 and before for.c 1.163 from 2022-01-09, 191# a newline character in a .for loop led to a crash since at the point where 192# the error message including the stack trace is printed, the body of the .for 193# loop is assembled, and at that point, ForLoop.nextItem had already been 194# advanced. 195.MAKEFLAGS: -dp 196.for i in "${.newline}" 197: $i 198.endfor 199.MAKEFLAGS: -d0 200 201.MAKEFLAGS: -df 202.for i in \# \\\# 203# $i 204.endfor 205 206.for i in $$ $$i $$(i) $${i} $$$$ $$$$$$$$ $${:U\$$\$$} 207# $i 208.endfor 209 210# The expression '${.TARGET}' must be preserved as it is one of the 7 built-in 211# target-local variables. See for.c 1.45 from 2009-01-14. 212.for i in ${.TARGET} $${.TARGET} $$${.TARGET} $$$${.TARGET} 213# $i 214.endfor 215# expect: # ${:U${.TARGET}} 216# XXX: Why does '$' result in the same text as '$$'? 217# expect: # ${:U${.TARGET}} 218# XXX: Why does the '$$' before the '${.TARGET}' lead to an escaped '}'? 219# expect: # ${:U$${.TARGET\}} 220# XXX: Why does '$' result in the same text as '$$'? 221# XXX: Why does the '$$' before the '${.TARGET}' lead to an escaped '}'? 222# expect: # ${:U$${.TARGET\}} 223 224.for i in ((( {{{ ))) }}} 225# $i 226.endfor 227.MAKEFLAGS: -d0 228 229all: 230