[bpt/coccinelle.git] / demos / unsigned.txt

From 12o3l@tiscali.nl Fri Apr 18 18:13:39 2008
Date: Fri, 18 Apr 2008 18:15:06 +0200
From: Roel Kluin <12o3l@tiscali.nl>
To: kernel-janitors@vger.kernel.org, kernelnewbies-bounce@nl.linux.org
Subject: script to find incorrect tests on unsigneds

A bash script to find incorrect tests on unsigned values. for instance:

unsigned int i;
...
i = neg_ret_function();
...
if (i < 0) ...

#!/bin/bash
# (c) roel kluin 2008 GPL v2
#
# TODO: make this working also for 
#                              ... $unsigned_var == $signed_var ...
#                              ... $unsigned_var == \(-POSITIVE_DEF\|NEGATIVE_DEF\) ...
#                              ... ( $unsigned_var [+*/%^&~-]?= ... ) < 0 ...

# a number
int="[0-9]"
hex="[a-f0-9]"
hEx="[A-Fa-f0-9]"
HEX="[A-F0-9]"
upp="[A-Z]"
up_="[A-Z_]"
low="[a-z0-9]"
lo_="[a-z0-9_]"
alp="[A-Za-z]"
al_="[A-Za-z_]"
ALN="[A-Z0-9]"
AN_="[A-Z0-9_]"
aln="[A-Za-z0-9]"
an_="[A-Za-z0-9_]"
# to match something like 1ul, floats or hexes as well:
D="$int*\.\?$int\+x\?$hex*[uUlL]\{0,3\}[fF]\?"

# more strict and catches it (costs one backreference for (git-)grep)
d="\($int\+[uUlLfF]\?\|$int\+[uU]\?[lL][lL]\?\|0x$hex\+\|0x$HEX\+\|$i\+[lL][lL][uU]\|$i*\.$i\+[fF]\?\)"

# capital: can be used to catch a definition or config option
K="$UP_\+$AN_*";

# can be used for a variable/function name:
V="$an_\+$an_*"

# works the same as above, but also for members and arrays: one backreference
# is more strict
W="$V\(\[$s$V$s\]\|\[$s$D$s\]\|\.$V\|->$V\)*"
# catches it at once (less strict)
w="\($V\|${V}\[$s$an_*${s}\]\|$V\.\|$V->\)\+"

# seperators:
s="[[:space:]]*";
S="[[:space:]]\+"

# useful to delimit a variable name:
Q="[^[:alnum:]_]"

# match the end of the line, including comments: one backreference (but at eol)
cendl="$s\(\/[\*\/].*\)\?$"

# match something that is not comment, string or character: 2 backreferences
ccode="\([^\/\"']*\|\/[^\*\/]\|\/\*\([^\*]*\|\**[^\/\*]\)*\*\+\/\|'[^']*'\|\"[^\"]*\"\)*"

# resp function open and end (only works when indentation is correct.
fo="^[\{]$cendl"
fe="^[\}]$cendl"
se="^[\}];$cendl"

# to match nested parentheses
nps="[^()]*"
nstdps="\(($nps\(($nps\(($nps\(($nps\(($nps)$nps\)*)$nps\)*)$nps\)*)$nps\)*)$nps\)*"


# first determine unsigned typedefs
arr="\(\[[^\]]*\]$s\)*"
attr="__attribute__$s(([^;]*))"
utype="${s}typedef${S}unsigned$S\($V$S\)*"
uns="unsigned$(
git-grep "^$utype\($V$s$arr\|$attr$S$V$s$arr\|$V$s$arr$S$attr\)$s;$cendl" | sed -n "s/^[^.]*\.[hc]:$utype\(\($V\)$s$arr\|$attr$S\($V\)$s$arr\|\($V\)$s$arr$S$attr\)$s;$cendl/\\\\|\3\5\7/p" | sort | uniq | tr -d "\n")"

# define left and right operators
# to decrease the number backrefences, these are assigned in loops
opl=
for op in "[;,|^?:(]" "[\!+*/%&|~^-]=" ">>=" "<<=" "\[" "&&" "$an_$s&"; do 
 opl="$opl$op\|$op$s++\|$op$s--\|"
done
opl="\(${opl%|})"
opr=
for op in "[;,&|^?:)]" "[\!+*/%&|~^<>-]=" ">>=" "<<=" ">[^>]" "<[^<]" "\]"; do 
 opr="$opr$op\|$op$s++\|$op$s--\|"
done
opr="\(${opr%|})"

# string catches invalid comparison
q1="$opl$s\($w$s\(>=${s}0\|<${s}0\|[><\!=]=$s-$s$D\|[<>]$s-$s$D\)\|\(0$s>\|0$s<=\|-$s$D${s}[><\!=]=\|-$s$D${s}[<>]\)$s$w\)$s$opr"

start=0
end=$(echo $uns | tr -cd "|" | wc -c)

# main function
while [ $start -lt $end ]; do
  # we match 30 typedefs at a time
  tuns="$(echo $uns | cut -d "\\" -f$start-$(($start+29)))"

  # catch candidate files 
  for f in $(git-grep -l "^\(${ccode}[,;]$s\)\?\($tuns\)$S" | grep "[^.]*\.[ch]" | xargs grep -l "$q1"); do
    for n in $(sed -n "/^.*$q1/=" $f); do # lines
      for v in $(sed -n "${n}s/^.*$q1.*$/\3\6/p" $f); do
        # n2 = wine there
        head -n$n $f | tac | sed -n "/^[{]/q; /^\(.*$Q\)\?\($tuns\)$S\($V$S\)*$v$Q.*$/=" | while read n2;        
        do
          echo "# --- invalid test on unsigned variable '$v' --- #"
          echo "vi $f +$n2 # unsigned declaration"
          echo "vi $f +$n # invalid test"
        done
      done
    done
  done
  start=$(($start+30))
done | less
--
To unsubscribe from this list: send the line "unsubscribe kernel-janitors" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Commit	Line	Data
34e49164 C	1	From 12o3l@tiscali.nl Fri Apr 18 18:13:39 2008
	2	Date: Fri, 18 Apr 2008 18:15:06 +0200
	3	From: Roel Kluin <12o3l@tiscali.nl>
	4	To: kernel-janitors@vger.kernel.org, kernelnewbies-bounce@nl.linux.org
	5	Subject: script to find incorrect tests on unsigneds
	6
	7	A bash script to find incorrect tests on unsigned values. for instance:
	8
	9	unsigned int i;
	10	...
	11	i = neg_ret_function();
	12	...
	13	if (i < 0) ...
	14
	15	#!/bin/bash
	16	# (c) roel kluin 2008 GPL v2
	17	#
	18	# TODO: make this working also for
	19	# ... $unsigned_var == $signed_var ...
	20	# ... $unsigned_var == \(-POSITIVE_DEF\\|NEGATIVE_DEF\) ...
	21	# ... ( $unsigned_var [+*/%^&~-]?= ... ) < 0 ...
	22
	23	# a number
	24	int="[0-9]"
	25	hex="[a-f0-9]"
	26	hEx="[A-Fa-f0-9]"
	27	HEX="[A-F0-9]"
	28	upp="[A-Z]"
	29	up_="[A-Z_]"
	30	low="[a-z0-9]"
	31	lo_="[a-z0-9_]"
	32	alp="[A-Za-z]"
	33	al_="[A-Za-z_]"
	34	ALN="[A-Z0-9]"
	35	AN_="[A-Z0-9_]"
	36	aln="[A-Za-z0-9]"
	37	an_="[A-Za-z0-9_]"
	38	# to match something like 1ul, floats or hexes as well:
	39	D="$int\.\?$int\+x\?$hex[uUlL]\{0,3\}[fF]\?"
	40
	41	# more strict and catches it (costs one backreference for (git-)grep)
	42	d="\($int\+[uUlLfF]\?\\|$int\+[uU]\?[lL][lL]\?\\|0x$hex\+\\|0x$HEX\+\\|$i\+[lL][lL][uU]\\|$i*\.$i\+[fF]\?\)"
	43
	44	# capital: can be used to catch a definition or config option
	45	K="$UP_\+$AN_*";
	46
	47	# can be used for a variable/function name:
	48	V="$an_\+$an_*"
	49
	50	# works the same as above, but also for members and arrays: one backreference
	51	# is more strict
	52	W="$V\(\[$s$V$s\]\\|\[$s$D$s\]\\|\.$V\\|->$V\)*"
	53	# catches it at once (less strict)
	54	w="\($V\\|${V}\[$s$an_*${s}\]\\|$V\.\\|$V->\)\+"
	55
	56	# seperators:
	57	s="[[:space:]]*";
	58	S="[[:space:]]\+"
	59
	60	# useful to delimit a variable name:
	61	Q="[^[:alnum:]_]"
	62
	63	# match the end of the line, including comments: one backreference (but at eol)
	64	cendl="$s\(\/[\\/].\)\?$"
65
66	# match something that is not comment, string or character: 2 backreferences
67	ccode="\([^\/\"']\\|\/[^\\/]\\|\/\\([^\]\\|\[^\/\]\)\\+\/\\|'[^']'\\|\"[^\"]\"\)*"
68
69	# resp function open and end (only works when indentation is correct.
70	fo="^[\{]$cendl"
71	fe="^[\}]$cendl"
72	se="^[\}];$cendl"
73
74	# to match nested parentheses
75	nps="[^()]*"
76	nstdps="\(($nps\(($nps\(($nps\(($nps\(($nps)$nps\))$nps\))$nps\))$nps\))$nps\)*"
77
78
79	# first determine unsigned typedefs
80	arr="\(\[[^\]]\]$s\)"
81	attr="__attribute__$s(([^;]*))"
82	utype="${s}typedef${S}unsigned$S\($V$S\)*"
83	uns="unsigned$(
84	git-grep "^$utype\($V$s$arr\\|$attr$S$V$s$arr\\|$V$s$arr$S$attr\)$s;$cendl" \| sed -n "s/^[^.]*\.[hc]:$utype\(\($V\)$s$arr\\|$attr$S\($V\)$s$arr\\|\($V\)$s$arr$S$attr\)$s;$cendl/\\\\\|\3\5\7/p" \| sort \| uniq \| tr -d "\n")"
85
86	# define left and right operators
87	# to decrease the number backrefences, these are assigned in loops
88	opl=
89	for op in "[;,\|^?:(]" "[\!+*/%&\|~^-]=" ">>=" "<<=" "\[" "&&" "$an_$s&"; do
90	opl="$opl$op\\|$op$s++\\|$op$s--\\|"
91	done
92	opl="\(${opl%\|})"
93	opr=
94	for op in "[;,&\|^?:)]" "[\!+*/%&\|~^<>-]=" ">>=" "<<=" ">[^>]" "<[^<]" "\]"; do
95	opr="$opr$op\\|$op$s++\\|$op$s--\\|"
96	done
97	opr="\(${opr%\|})"
98
99	# string catches invalid comparison
100	q1="$opl$s\($w$s\(>=${s}0\\|<${s}0\\|[><\!=]=$s-$s$D\\|[<>]$s-$s$D\)\\|\(0$s>\\|0$s<=\\|-$s$D${s}[><\!=]=\\|-$s$D${s}[<>]\)$s$w\)$s$opr"
101
102	start=0
103	end=$(echo $uns \| tr -cd "\|" \| wc -c)
104
105	# main function
106	while [ $start -lt $end ]; do
107	# we match 30 typedefs at a time
108	tuns="$(echo $uns \| cut -d "\\" -f$start-$(($start+29)))"
109
110	# catch candidate files
111	for f in $(git-grep -l "^\(${ccode}[,;]$s\)\?\($tuns\)$S" \| grep "[^.]*\.[ch]" \| xargs grep -l "$q1"); do
112	for n in $(sed -n "/^.*$q1/=" $f); do # lines
113	for v in $(sed -n "${n}s/^.$q1.$/\3\6/p" $f); do
114	# n2 = wine there
115	head -n$n $f \| tac \| sed -n "/^[{]/q; /^\(.$Q\)\?\($tuns\)$S\($V$S\)$v$Q.*$/=" \| while read n2;
116	do
117	echo "# --- invalid test on unsigned variable '$v' --- #"
118	echo "vi $f +$n2 # unsigned declaration"
119	echo "vi $f +$n # invalid test"
120	done
121	done
122	done
123	done
124	start=$(($start+30))
125	done \| less
126	--
127	To unsubscribe from this list: send the line "unsubscribe kernel-janitors" in
128	the body of a message to majordomo@vger.kernel.org
129	More majordomo info at http://vger.kernel.org/majordomo-info.html