| @@ -1,167 +1,172 @@ | | | @@ -1,167 +1,172 @@ |
1 | # $NetBSD: varmod-edge.mk,v 1.9 2020/08/01 15:13:45 rillig Exp $ | | 1 | # $NetBSD: varmod-edge.mk,v 1.10 2020/08/01 15:16:15 rillig Exp $ |
2 | # | | 2 | # |
3 | # Tests for edge cases in variable modifiers. | | 3 | # Tests for edge cases in variable modifiers. |
4 | # | | 4 | # |
5 | # These tests demonstrate the current implementation in small examples. | | 5 | # These tests demonstrate the current implementation in small examples. |
6 | # They may contain surprising behavior. | | 6 | # They may contain surprising behavior. |
7 | # | | 7 | # |
8 | # Each test consists of: | | 8 | # Each test consists of: |
9 | # - INP, the input to the test | | 9 | # - INP, the input to the test |
10 | # - MOD, the expression for testing the modifier | | 10 | # - MOD, the expression for testing the modifier |
11 | # - EXP, the expected output | | 11 | # - EXP, the expected output |
12 | | | 12 | |
13 | TESTS+= M-paren | | 13 | TESTS+= M-paren |
14 | INP.M-paren= (parentheses) {braces} (opening closing) () | | 14 | INP.M-paren= (parentheses) {braces} (opening closing) () |
15 | MOD.M-paren= ${INP.M-paren:M(*)} | | 15 | MOD.M-paren= ${INP.M-paren:M(*)} |
16 | EXP.M-paren= (parentheses) () | | 16 | EXP.M-paren= (parentheses) () |
17 | | | 17 | |
18 | # The first closing brace matches the opening parenthesis. | | 18 | # The first closing brace matches the opening parenthesis. |
19 | # The second closing brace actually ends the variable expression. | | 19 | # The second closing brace actually ends the variable expression. |
20 | # | | 20 | # |
21 | # XXX: This is unexpected but rarely occurs in practice. | | 21 | # XXX: This is unexpected but rarely occurs in practice. |
22 | TESTS+= M-mixed | | 22 | TESTS+= M-mixed |
23 | INP.M-mixed= (paren-brace} ( | | 23 | INP.M-mixed= (paren-brace} ( |
24 | MOD.M-mixed= ${INP.M-mixed:M(*}} | | 24 | MOD.M-mixed= ${INP.M-mixed:M(*}} |
25 | EXP.M-mixed= (paren-brace} | | 25 | EXP.M-mixed= (paren-brace} |
26 | | | 26 | |
27 | # After the :M modifier has parsed the pattern, only the closing brace | | 27 | # After the :M modifier has parsed the pattern, only the closing brace |
28 | # and the colon are unescaped. The other characters are left as-is. | | 28 | # and the colon are unescaped. The other characters are left as-is. |
29 | # To actually see this effect, the backslashes in the :M modifier need | | 29 | # To actually see this effect, the backslashes in the :M modifier need |
30 | # to be doubled since single backslashes would simply be unescaped by | | 30 | # to be doubled since single backslashes would simply be unescaped by |
31 | # Str_Match. | | 31 | # Str_Match. |
32 | # | | 32 | # |
33 | # XXX: This is unexpected. The opening brace should also be unescaped. | | 33 | # XXX: This is unexpected. The opening brace should also be unescaped. |
34 | TESTS+= M-unescape | | 34 | TESTS+= M-unescape |
35 | INP.M-unescape= ({}): \(\{\}\)\: \(\{}\): | | 35 | INP.M-unescape= ({}): \(\{\}\)\: \(\{}\): |
36 | MOD.M-unescape= ${INP.M-unescape:M\\(\\{\\}\\)\\:} | | 36 | MOD.M-unescape= ${INP.M-unescape:M\\(\\{\\}\\)\\:} |
37 | EXP.M-unescape= \(\{}\): | | 37 | EXP.M-unescape= \(\{}\): |
38 | | | 38 | |
39 | # When the :M and :N modifiers are parsed, the pattern finishes as soon | | 39 | # When the :M and :N modifiers are parsed, the pattern finishes as soon |
40 | # as open_parens + open_braces == closing_parens + closing_braces. This | | 40 | # as open_parens + open_braces == closing_parens + closing_braces. This |
41 | # means that ( and } form a matching pair. | | 41 | # means that ( and } form a matching pair. |
42 | # | | 42 | # |
43 | # Nested variable expressions are not parsed as such. Instead, only the | | 43 | # Nested variable expressions are not parsed as such. Instead, only the |
44 | # parentheses and braces are counted. This leads to a parse error since | | 44 | # parentheses and braces are counted. This leads to a parse error since |
45 | # the nested expression is not "${:U*)}" but only "${:U*)", which is | | 45 | # the nested expression is not "${:U*)}" but only "${:U*)", which is |
46 | # missing the closing brace. The expression is evaluated anyway. | | 46 | # missing the closing brace. The expression is evaluated anyway. |
47 | # The final brace in the output comes from the end of M.nest-mix. | | 47 | # The final brace in the output comes from the end of M.nest-mix. |
48 | # | | 48 | # |
49 | # XXX: This is unexpected but rarely occurs in practice. | | 49 | # XXX: This is unexpected but rarely occurs in practice. |
50 | TESTS+= M-nest-mix | | 50 | TESTS+= M-nest-mix |
51 | INP.M-nest-mix= (parentheses) | | 51 | INP.M-nest-mix= (parentheses) |
52 | MOD.M-nest-mix= ${INP.M-nest-mix:M${:U*)}} | | 52 | MOD.M-nest-mix= ${INP.M-nest-mix:M${:U*)}} |
53 | EXP.M-nest-mix= (parentheses)} | | 53 | EXP.M-nest-mix= (parentheses)} |
54 | # make: Unclosed variable specification (expecting '}') for "" (value "*)") modifier U | | 54 | # make: Unclosed variable specification (expecting '}') for "" (value "*)") modifier U |
55 | | | 55 | |
56 | # In contrast to parentheses and braces, the brackets are not counted | | 56 | # In contrast to parentheses and braces, the brackets are not counted |
57 | # when the :M modifier is parsed since Makefile variables only take the | | 57 | # when the :M modifier is parsed since Makefile variables only take the |
58 | # ${VAR} or $(VAR) forms, but not $[VAR]. | | 58 | # ${VAR} or $(VAR) forms, but not $[VAR]. |
59 | # | | 59 | # |
60 | # The final ] in the pattern is needed to close the character class. | | 60 | # The final ] in the pattern is needed to close the character class. |
61 | TESTS+= M-nest-brk | | 61 | TESTS+= M-nest-brk |
62 | INP.M-nest-brk= [ [[ [[[ | | 62 | INP.M-nest-brk= [ [[ [[[ |
63 | MOD.M-nest-brk= ${INP.M-nest-brk:M${:U[[[[[]}} | | 63 | MOD.M-nest-brk= ${INP.M-nest-brk:M${:U[[[[[]}} |
64 | EXP.M-nest-brk= [ | | 64 | EXP.M-nest-brk= [ |
65 | | | 65 | |
66 | # The pattern in the nested variable has an unclosed character class. | | 66 | # The pattern in the nested variable has an unclosed character class. |
67 | # No error is reported though, and the pattern is closed implicitly. | | 67 | # No error is reported though, and the pattern is closed implicitly. |
68 | # | | 68 | # |
69 | # XXX: It is unexpected that no error is reported. | | 69 | # XXX: It is unexpected that no error is reported. |
70 | # See str.c, function Str_Match. | | 70 | # See str.c, function Str_Match. |
71 | # | | 71 | # |
72 | # Before 2019-12-02, this test case triggered an out-of-bounds read | | 72 | # Before 2019-12-02, this test case triggered an out-of-bounds read |
73 | # in Str_Match. | | 73 | # in Str_Match. |
74 | TESTS+= M-pat-err | | 74 | TESTS+= M-pat-err |
75 | INP.M-pat-err= [ [[ [[[ | | 75 | INP.M-pat-err= [ [[ [[[ |
76 | MOD.M-pat-err= ${INP.M-pat-err:M${:U[[}} | | 76 | MOD.M-pat-err= ${INP.M-pat-err:M${:U[[}} |
77 | EXP.M-pat-err= [ | | 77 | EXP.M-pat-err= [ |
78 | | | 78 | |
79 | # The first backslash does not escape the second backslash. | | 79 | # The first backslash does not escape the second backslash. |
80 | # Therefore, the second backslash escapes the parenthesis. | | 80 | # Therefore, the second backslash escapes the parenthesis. |
81 | # This means that the pattern ends there. | | 81 | # This means that the pattern ends there. |
82 | # The final } in the output comes from the end of MOD.M-bsbs. | | 82 | # The final } in the output comes from the end of MOD.M-bsbs. |
83 | # | | 83 | # |
84 | # If the first backslash were to escape the second backslash, the first | | 84 | # If the first backslash were to escape the second backslash, the first |
85 | # closing brace would match the opening parenthesis (see M-mixed), and | | 85 | # closing brace would match the opening parenthesis (see M-mixed), and |
86 | # the second closing brace would be needed to close the variable. | | 86 | # the second closing brace would be needed to close the variable. |
87 | # After that, the remaining backslash would escape the parenthesis in | | 87 | # After that, the remaining backslash would escape the parenthesis in |
88 | # the pattern, therefore (} would match. | | 88 | # the pattern, therefore (} would match. |
89 | TESTS+= M-bsbs | | 89 | TESTS+= M-bsbs |
90 | INP.M-bsbs= (} \( \(} | | 90 | INP.M-bsbs= (} \( \(} |
91 | MOD.M-bsbs= ${INP.M-bsbs:M\\(}} | | 91 | MOD.M-bsbs= ${INP.M-bsbs:M\\(}} |
92 | EXP.M-bsbs= \(} | | 92 | EXP.M-bsbs= \(} |
93 | #EXP.M-bsbs= (} # If the first backslash were to escape ... | | 93 | #EXP.M-bsbs= (} # If the first backslash were to escape ... |
94 | | | 94 | |
95 | # The backslash in \( does not escape the parenthesis, therefore it | | 95 | # The backslash in \( does not escape the parenthesis, therefore it |
96 | # counts for the nesting level and matches with the first closing brace. | | 96 | # counts for the nesting level and matches with the first closing brace. |
97 | # The second closing brace closes the variable, and the third is copied | | 97 | # The second closing brace closes the variable, and the third is copied |
98 | # literally. | | 98 | # literally. |
99 | # | | 99 | # |
100 | # The second :M in the pattern is nested between ( and }, therefore it | | 100 | # The second :M in the pattern is nested between ( and }, therefore it |
101 | # does not start a new modifier. | | 101 | # does not start a new modifier. |
102 | TESTS+= M-bs1-par | | 102 | TESTS+= M-bs1-par |
103 | INP.M-bs1-par= ( (:M (:M} \( \(:M \(:M} | | 103 | INP.M-bs1-par= ( (:M (:M} \( \(:M \(:M} |
104 | MOD.M-bs1-par= ${INP.M-bs1-par:M\(:M*}}} | | 104 | MOD.M-bs1-par= ${INP.M-bs1-par:M\(:M*}}} |
105 | EXP.M-bs1-par= (:M}} | | 105 | EXP.M-bs1-par= (:M}} |
106 | | | 106 | |
107 | # The double backslash is passed verbatim to the pattern matcher. | | 107 | # The double backslash is passed verbatim to the pattern matcher. |
108 | # The Str_Match pattern is \\(:M*}, and there the backslash is unescaped. | | 108 | # The Str_Match pattern is \\(:M*}, and there the backslash is unescaped. |
109 | # Again, the ( takes place in the nesting level, and there is no way to | | 109 | # Again, the ( takes place in the nesting level, and there is no way to |
110 | # prevent this, no matter how many backslashes are used. | | 110 | # prevent this, no matter how many backslashes are used. |
111 | TESTS+= M-bs2-par | | 111 | TESTS+= M-bs2-par |
112 | INP.M-bs2-par= ( (:M (:M} \( \(:M \(:M} | | 112 | INP.M-bs2-par= ( (:M (:M} \( \(:M \(:M} |
113 | MOD.M-bs2-par= ${INP.M-bs2-par:M\\(:M*}}} | | 113 | MOD.M-bs2-par= ${INP.M-bs2-par:M\\(:M*}}} |
114 | EXP.M-bs2-par= \(:M}} | | 114 | EXP.M-bs2-par= \(:M}} |
115 | | | 115 | |
116 | # Str_Match uses a recursive algorithm for matching the * patterns. | | 116 | # Str_Match uses a recursive algorithm for matching the * patterns. |
117 | # Make sure that it survives patterns with 128 asterisks. | | 117 | # Make sure that it survives patterns with 128 asterisks. |
118 | # That should be enough for all practical purposes. | | 118 | # That should be enough for all practical purposes. |
119 | # To produce a stack overflow, just add more :Qs below. | | 119 | # To produce a stack overflow, just add more :Qs below. |
120 | TESTS+= M-128 | | 120 | TESTS+= M-128 |
121 | INP.M-128= ${:U\\:Q:Q:Q:Q:Q:Q:Q:S,\\,x,g} | | 121 | INP.M-128= ${:U\\:Q:Q:Q:Q:Q:Q:Q:S,\\,x,g} |
122 | PAT.M-128= ${:U\\:Q:Q:Q:Q:Q:Q:Q:S,\\,*,g} | | 122 | PAT.M-128= ${:U\\:Q:Q:Q:Q:Q:Q:Q:S,\\,*,g} |
123 | MOD.M-128= ${INP.M-128:M${PAT.M-128}} | | 123 | MOD.M-128= ${INP.M-128:M${PAT.M-128}} |
124 | EXP.M-128= ${INP.M-128} | | 124 | EXP.M-128= ${INP.M-128} |
125 | | | 125 | |
126 | # This is the normal SysV substitution. Nothing surprising here. | | 126 | # This is the normal SysV substitution. Nothing surprising here. |
127 | TESTS+= eq-ext | | 127 | TESTS+= eq-ext |
128 | INP.eq-ext= file.c file.cc | | 128 | INP.eq-ext= file.c file.cc |
129 | MOD.eq-ext= ${INP.eq-ext:%.c=%.o} | | 129 | MOD.eq-ext= ${INP.eq-ext:%.c=%.o} |
130 | EXP.eq-ext= file.o file.cc | | 130 | EXP.eq-ext= file.o file.cc |
131 | | | 131 | |
132 | # The SysV := modifier is greedy and consumes all the modifier text | | 132 | # The SysV := modifier is greedy and consumes all the modifier text |
133 | # up until the closing brace or parenthesis. The :Q may look like a | | 133 | # up until the closing brace or parenthesis. The :Q may look like a |
134 | # modifier, but it really isn't, that's why it appears in the output. | | 134 | # modifier, but it really isn't, that's why it appears in the output. |
135 | TESTS+= eq-q | | 135 | TESTS+= eq-q |
136 | INP.eq-q= file.c file.cc | | 136 | INP.eq-q= file.c file.cc |
137 | MOD.eq-q= ${INP.eq-q:%.c=%.o:Q} | | 137 | MOD.eq-q= ${INP.eq-q:%.c=%.o:Q} |
138 | EXP.eq-q= file.o:Q file.cc | | 138 | EXP.eq-q= file.o:Q file.cc |
139 | | | 139 | |
140 | # The = in the := modifier can be escaped. | | 140 | # The = in the := modifier can be escaped. |
141 | TESTS+= eq-bs | | 141 | TESTS+= eq-bs |
142 | INP.eq-bs= file.c file.c=%.o | | 142 | INP.eq-bs= file.c file.c=%.o |
143 | MOD.eq-bs= ${INP.eq-bs:%.c\=%.o=%.ext} | | 143 | MOD.eq-bs= ${INP.eq-bs:%.c\=%.o=%.ext} |
144 | EXP.eq-bs= file.c file.ext | | 144 | EXP.eq-bs= file.c file.ext |
145 | | | 145 | |
146 | # Having only an escaped '=' results in a parse error. | | 146 | # Having only an escaped '=' results in a parse error. |
147 | # The call to "pattern.lhs = ParseModifierPart" fails. | | 147 | # The call to "pattern.lhs = ParseModifierPart" fails. |
148 | TESTS+= eq-esc | | 148 | TESTS+= eq-esc |
149 | INP.eq-esc= file.c file... | | 149 | INP.eq-esc= file.c file... |
150 | MOD.eq-esc= ${INP.eq-esc:a\=b} | | 150 | MOD.eq-esc= ${INP.eq-esc:a\=b} |
151 | EXP.eq-esc= # empty | | 151 | EXP.eq-esc= # empty |
152 | # make: Unclosed substitution for INP.eq-esc (= missing) | | 152 | # make: Unclosed substitution for INP.eq-esc (= missing) |
153 | | | 153 | |
154 | TESTS+= colon | | 154 | TESTS+= colon |
155 | INP.colon= value | | 155 | INP.colon= value |
156 | MOD.colon= ${INP.colon:} | | 156 | MOD.colon= ${INP.colon:} |
157 | EXP.colon= value | | 157 | EXP.colon= value |
158 | | | 158 | |
| | | 159 | TESTS+= colons |
| | | 160 | INP.colons= value |
| | | 161 | MOD.colons= ${INP.colons::::} |
| | | 162 | EXP.colons= # empty |
| | | 163 | |
159 | all: | | 164 | all: |
160 | .for test in ${TESTS} | | 165 | .for test in ${TESTS} |
161 | . if ${MOD.${test}} == ${EXP.${test}} | | 166 | . if ${MOD.${test}} == ${EXP.${test}} |
162 | @printf 'ok %s\n' ${test:Q}'' | | 167 | @printf 'ok %s\n' ${test:Q}'' |
163 | . else | | 168 | . else |
164 | @printf 'error in %s: expected %s, got %s\n' \ | | 169 | @printf 'error in %s: expected %s, got %s\n' \ |
165 | ${test:Q}'' ${EXP.${test}:Q}'' ${MOD.${test}:Q}'' | | 170 | ${test:Q}'' ${EXP.${test}:Q}'' ${MOD.${test}:Q}'' |
166 | . endif | | 171 | . endif |
167 | .endfor | | 172 | .endfor |