Fix multiple long-standing bugs, improve test suite.
diff --git a/ChangeLog b/ChangeLog
new file mode 100644
index 0000000..17715fc
--- /dev/null
+++ b/ChangeLog
@@ -0,0 +1,225 @@
+2018-08-22 Arnold D. Robbins <arnold@skeeve.com>
+
+ * awktest.tar (testdir/T.expr): Fix test for unary plus.
+
+2018-08-22 Arnold D. Robbins <arnold@skeeve.com>
+
+ * REGRESS: Extract tests if necessary, set PATH to include '.'.
+ * regdir/beebe.tar (Makefile): Fix longwrds test to prefix
+ sort with LC_ALL=C.
+ * awktest.tar: Updated from fixed test suite, directory
+ it extracts is now called 'testdir' to match what's in top-level
+ REGRESS script.
+ * regdir: Removed, as Brian wants to keep the test suite in
+ the tar file.
+
+2018-08-22 Arnold D. Robbins <arnold@skeeve.com>
+
+ * FIXES, lib.c, run.c, makefile, main.c: Merge from Brian's tree.
+ * REGRESS: New file, from Brian.
+ * awktest.tar: Restored from Brian's tree.
+
+2018-08-22 Arnold D. Robbins <arnold@skeeve.com>
+
+ * awkgram.y (UPLUS): New token. In the grammar, call op1()
+ with it.
+ * maketab.c (proc): Add entry for UPLUS.
+ * run.c (arith): Handle UPLUS.
+ * main.c (version): Updated.
+ * bugs-fixed/unary-plus.awk, bugs-fixed/unary-plus.bad,
+ bugs-fixed/unary-plus.ok: New files.
+
+2018-08-10 Arnold D. Robbins <arnold@skeeve.com>
+
+ * TODO: Updated.
+ * awk.1: Improve use of macros, add some additional explanation
+ in a few places, alphabetize list of variables.
+
+2018-08-08 Arnold D. Robbins <arnold@skeeve.com>
+
+ * awk.h (Cell): Add new field `fmt' to track xFMT value used
+ for a string conversion.
+ [CONVC, CONVO]: New flag macros.
+ * bugs-fixed/README: Updated.
+ * bugs-fixed/string-conv.awk, bugs-fixed/string-conv.bad,
+ bugs-fixed/string-conv.ok: New files.
+ * main.c (version): Updated.
+ * proto.h (flags2str): Add declaration.
+ * tran.c (setfval): Clear CONVC and CONVO flags and set vp->fmt
+ to NULL.
+ (setsval): Ditto. Add large comment and new code to manage
+ correct conversion of number to string based on various flags
+ and the value of vp->fmt. The idea is to not convert again
+ if xFMT is the same as before and we're doing the same conversion.
+ Otherwise, clear the old flags, set the new, and reconvert.
+ (flags2str): New function. For debug prints and for use from a debugger.
+
+2018-08-05 Arnold D. Robbins <arnold@skeeve.com>
+
+ Fix filename conflicts in regdir where the only difference was
+ in letter case. This caused problems on Windows systems.
+
+ * regdir/Compare.T1: Renamed from regdir/Compare.T.
+ * regdir/t.delete0: Renamed from regdir/t.delete.
+ * regdir/t.getline1: Renamed from regdir/t.getline.
+ * regdir/t.redir1: Renamed from regdir/t.redir.
+ * regdir/t.split1: Renamed from regdir/t.split.
+ * regdir/t.sub0: Renamed from regdir/t.sub.
+ * regdir/REGRESS: Adjusted.
+
+2018-08-04 Arnold D. Robbins <arnold@skeeve.com>
+
+ With scalpel, tweasers, magnifying glass and bated breath,
+ borrow code from the NetBSD version of nawk to fix the years-old
+ bug whereby decrementing the value of NF did not change the
+ record.
+
+ * lib.c (fldbld): Set donerec to 1 when done.
+ (setlastfld): New function.
+ * proto.h (setlastfld): Add declaration.
+ * run.c (copycell): Make code smarter about flags (from NetBSD code).
+ * tran.c (setfree): New function.
+ * tran.c (setfval): Normalize negative zero to positive zero.
+ If setting NF, clear donerec and call setlastfld().
+ (setsval): Remove call to save_old_OFS(). If setting OFS, call
+ recbld(). If setting NF, clear donerec and call setlastfld().
+
+ As part of the process, revert OFS-related changes of 2018-05-22:
+
+ * awk.h (saveOFS, saveOFSlen, save_old_OFS): Remove declarations.
+ * lib.c (recbld): Use *OFS instead of saveOFS.
+ * run.c (saveOFS, saveOFSlen, save_old_OFS): Remove.
+ * tran.c (syminit): Remove initialization of saveOFS and saveOFSlen.
+
+ General stuff that goes along with all this:
+
+ * bugs-fixed/README: Updated.
+ * bugs-fixed/decr-NF.awk, bugs-fixed/decr-NF.bad,
+ bugs-fixed/decr-NF.ok: New files.
+ * main.c (version): Updated.
+ * regdir/README.TESTS: Fix awk book title.
+ * regdir/T.misc: Revise test to match fixed code.
+ * run.c (format): Increase size of buffer used for %a test. (Unrelated
+ to NF or OFS, but fixes a compiler complaint.)
+
+2018-06-07 Arnold D. Robbins <arnold@skeeve.com>
+
+ * regdir/beebe.tar: Fix longwrds.ok so that the test will pass.
+ The file was incorrectly sorted.
+
+2018-06-06 Arnold D. Robbins <arnold@skeeve.com>
+
+ * regdir/T.lilly: Fix the bug again in the second instance
+ of the code. Thanks to BWK for pointing this out.
+
+2018-05-31 Arnold D. Robbins <arnold@skeeve.com>
+
+ * regdir/T.lilly: Fix a syntax error and ordering bug
+ in creating the 'foo' file.
+
+2018-05-23 Arnold D. Robbins <arnold@skeeve.com>
+
+ * awk.1: Remove standalone 'awk' at the top of file, it messed up
+ the formatting. Arrange built-in variable list in alphabetical
+ order.
+
+2018-05-23 Arnold D. Robbins <arnold@skeeve.com>
+
+ * main.c (version): Add my email address and a date so that
+ users can tell this isn't straight BWK awk.
+ * README.md: Minor updates.
+ * TODO: Updated.
+
+2018-05-22 Arnold D. Robbins <arnold@skeeve.com>
+
+ Add POSIX-required formats %a and %A.
+
+ * run.c (format): Check for %a support in C library. If there,
+ allow %a and %A as valid formats.
+ * TODO: Updated.
+ * bugs-fixed/README: Updated.
+ * bugs-fixed/a-format.awk, bugs-fixed/a-format.bad,
+ bugs-fixed/a-format.ok: New files.
+
+2018-05-22 Arnold D. Robbins <arnold@skeeve.com>
+
+ * FIXES: Restored a line from a much earlier version that
+ apparently got lost when the dates were reordered.
+ * TODO: Updated.
+
+2018-05-22 Arnold D. Robbins <arnold@skeeve.com>
+
+ * README.md: New file.
+
+2018-05-22 Arnold D. Robbins <arnold@skeeve.com>
+
+ * regdir/echo.c, regdir/time.c: Minor fixes to compile without
+ warning on current GCC / Linux.
+
+2018-05-22 Arnold D. Robbins <arnold@skeeve.com>
+
+ * TODO: New file.
+
+2018-05-22 Arnold D. Robbins <arnold@skeeve.com>
+
+ * makefile (gitadd, gitpush): Remove these targets. They
+ should not be automated and were incorrect for things that
+ would be done regularly.
+
+2018-05-22 Arnold D. Robbins <arnold@skeeve.com>
+
+ Fix nawk so that [[:blank:]] only matches space and tab instead
+ of any whitespace character, originally made May 10, 2018.
+ See bugs-fixed/space.awk.
+
+ This appears to have been a thinko on Brian's part.
+
+ * b.c (charclasses): Use xisblank() function for [[:blank:]].
+ * bugs-fixed/README: Updated.
+ * bugs-fixed/space.awk, bugs-fixed/space.bad,
+ bugs-fixed/space.ok: New files.
+
+2018-05-22 Arnold D. Robbins <arnold@skeeve.com>
+
+ * .gitignore: New file.
+
+2018-05-22 Arnold D. Robbins <arnold@skeeve.com>
+
+ Fix nawk to provide reasonable exit status for system(),
+ a la gawk, originally made March 12, 2016. See
+ bugs-fixed/system-status.awk.
+
+ * run.c (bltin): For FSYSTEM, use the macros defined for wait(2)
+ to produce a reasonable exit value, instead of doing a floating-point
+ division by 256.
+ * awk.1: Document the return status values.
+ * bugs-fixed/README: Updated.
+ * bugs-fixed/system-status.awk, bugs-fixed/system-status.bad,
+ bugs-fixed/system-status.ok: New files.
+
+2018-05-22 Arnold D. Robbins <arnold@skeeve.com>
+
+ Bug fix with respect to rebuilding a record, originally
+ made August 19, 2014. See bugs-fixed/ofs-rebuild.awk.
+
+ * awk.h (saveOFS, saveOFSlen): Declare new variables.
+ * lib.c (recbld): Use them when rebuilding the record.
+ * run.c (saveOFS, saveOFSlen): Define new variables.
+ (save_old_OFS): New function to save OFS aside.
+ * tran.c (syminit): Initialize saveOFS and saveOFSlen.
+ (setsval): If setting a field, call save_old_OFS().
+ * bugs-fixed/README, bugs-fixed/ofs-rebuild.awk,
+ bugs-fixed/ofs-rebuild.bad, bugs-fixed/ofs-rebuild.ok: New files.
+
+2018-05-22 Arnold D. Robbins <arnold@skeeve.com>
+
+ * makefile (YACC): Use bison.
+
+2018-05-22 Arnold D. Robbins <arnold@skeeve.com>
+
+ * ChangeLog: Created.
+ * regdir: Created. Based on contents of awktest.a.
+ * .gitattributes: Created, to preserve CR LF in regdir/t.crlf.
+ * awktest.a: Removed.
+ * regdir/T.gawk, regdir/T.latin1: Updated from awktest.tar.
+ * awktest.tar: Removed.
diff --git a/FIXES b/FIXES
index b17d8a3..5c33dd9 100644
--- a/FIXES
+++ b/FIXES
@@ -514,6 +514,8 @@
Mar 12, 1998:
added -V to print version number and die.
+[notify dave kerns, dkerns@dacsoup.ih.lucent.com]
+
Feb 11, 1998:
subtle silent bug in lex.c: if the program ended with a number
longer than 1 digit, part of the input would be pushed back and
diff --git a/REGRESS b/REGRESS
index f4372d6..facbd83 100755
--- a/REGRESS
+++ b/REGRESS
@@ -1,3 +1,20 @@
+#! /bin/sh
+
+if [ -d testdir ]
+then
+ true # do nothing
+elif [ -f awktest.tar ]
+then
+ echo extracting testdir
+ tar -xpf awktest.tar
+else
+ echo $0: No testdir directory and no awktest.tar to extract it from! >&2
+ exit 1
+fi
+
cd testdir
pwd
+PATH=.:$PATH
+export PATH
+
REGRESS
diff --git a/TODO b/TODO
new file mode 100644
index 0000000..5099569
--- /dev/null
+++ b/TODO
@@ -0,0 +1,20 @@
+Fri Aug 10 11:11:11 IDT 2018
+============================
+
+A semi-random list of things to look into.
+
+1. DONE. Make sure the FIXES file has everything after the date order
+was reversed.
+
+2. DONE. Look into the problem that NF-- doesn't throw away the last field.
+This a long-standing issue.
+
+3. DONE. Add suport for %a and %A printf formats.
+
+4. DONE. Look into the problem that string conversions are sticky; once
+done they never change, even if OFMT or CONVFMT do. Similarly, a
+conversion from OFMT becomes the permanent string value, which is
+incorrect.
+
+5. MAYBE: Improve the test suite enough so that it can be cleaned up after
+it was run, and old results can be saved for comparison.
diff --git a/awk.1 b/awk.1
index 6119613..5830143 100644
--- a/awk.1
+++ b/awk.1
@@ -7,7 +7,6 @@
.fi
.ft 1
..
-awk
.TH AWK 1
.CT 1 files prog_other
.SH NAME
@@ -36,7 +35,7 @@
scans each input
.I file
for lines that match any of a set of patterns specified literally in
-.IR prog
+.I prog
or in one or more files
specified as
.B \-f
@@ -53,7 +52,7 @@
.B \-
means the standard input.
Any
-.IR file
+.I file
of the form
.I var=value
is treated as an assignment, not a filename,
@@ -70,12 +69,12 @@
options may be present.
The
.B \-F
-.IR fs
+.I fs
option defines the input field separator to be the regular expression
-.IR fs.
+.IR fs .
.PP
An input line is normally made up of fields separated by white space,
-or by regular expression
+or by the regular expression
.BR FS .
The fields are denoted
.BR $1 ,
@@ -87,7 +86,7 @@
.BR FS
is null, the input line is split into one field per character.
.PP
-A pattern-action statement has the form
+A pattern-action statement has the form:
.IP
.IB pattern " { " action " }
.PP
@@ -101,7 +100,7 @@
A statement can be one of the following:
.PP
.EX
-.ta \w'\f(CWdelete array[expression]'u
+.ta \w'\f(CWdelete array[expression]\fR'u
.RS
.nf
.ft CW
@@ -145,7 +144,7 @@
are also available in expressions.
Variables may be scalars, array elements
(denoted
-.IB x [ i ] )
+.IB x [ i ] \fR)
or fields.
Variables are initialized to the null string.
Array subscripts may be any string,
@@ -161,11 +160,11 @@
.B print
statement prints its arguments on the standard output
(or on a file if
-.BI > file
+.BI > " file
or
-.BI >> file
+.BI >> " file
is present or on a pipe if
-.BI | cmd
+.BI | " cmd
is present), separated by the current output field separator,
and terminated by the output record separator.
.I file
@@ -176,9 +175,10 @@
the same open file.
The
.B printf
-statement formats its expression list according to the format
+statement formats its expression list according to the
+.I format
(see
-.IR printf (3)) .
+.IR printf (3)).
The built-in function
.BI close( expr )
closes the file or pipe
@@ -189,13 +189,13 @@
.IR expr .
.PP
The mathematical functions
+.BR atan2 ,
+.BR cos ,
.BR exp ,
.BR log ,
-.BR sqrt ,
.BR sin ,
-.BR cos ,
and
-.BR atan2
+.B sqrt
are built in.
Other built-in functions:
.TF length
@@ -203,7 +203,8 @@
.B length
the length of its argument
taken as a string,
-or of
+number of elements in an array for an array argument,
+or length of
.B $0
if no argument.
.TP
@@ -218,14 +219,18 @@
.B int
truncates to an integer value
.TP
-.BI substr( s , " m" , " n\fB)
+\fBsubstr(\fIs\fB, \fIm\fR [\fB, \fIn\^\fR]\fB)\fR
the
.IR n -character
substring of
.I s
that begins at position
-.IR m
+.I m
counted from 1.
+If no
+.IR m ,
+use the rest of the string
+.I
.TP
.BI index( s , " t" )
the position in
@@ -246,14 +251,14 @@
.B RLENGTH
are set to the position and length of the matched string.
.TP
-.BI split( s , " a" , " fs\fB)
+\fBsplit(\fIs\fB, \fIa \fR[\fB, \fIfs\^\fR]\fB)\fR
splits the string
.I s
into array elements
-.IB a [1] ,
-.IB a [2] ,
+.IB a [1] \fR,
+.IB a [2] \fR,
\&...,
-.IB a [ n ] ,
+.IB a [ n ] \fR,
and returns
.IR n .
The separation is done with the regular expression
@@ -266,7 +271,7 @@
An empty string as field separator splits the string
into one array element per character.
.TP
-.BI sub( r , " t" , " s\fB)
+\fBsub(\fIr\fB, \fIt \fR[, \fIs\^\fR]\fB)
substitutes
.I t
for the first occurrence of the regular expression
@@ -279,7 +284,7 @@
.B $0
is used.
.TP
-.B gsub
+\fBgsub(\fIr\fB, \fIt \fR[, \fIs\^\fR]\fB)
same as
.B sub
except that all occurrences of the regular expression
@@ -289,18 +294,28 @@
.B gsub
return the number of replacements.
.TP
-.BI sprintf( fmt , " expr" , " ...\fB )
+.BI sprintf( fmt , " expr" , " ...\fB)
the string resulting from formatting
.I expr ...
according to the
.IR printf (3)
format
-.I fmt
+.IR fmt .
.TP
.BI system( cmd )
executes
.I cmd
-and returns its exit status
+and returns its exit status. This will be \-1 upon error,
+.IR cmd 's
+exit status upon a normal exit,
+256 +
+.I sig
+upon death-by-signal, where
+.I sig
+is the number of the murdering signal,
+or 512 +
+.I sig
+if there was a core dump.
.TP
.BI tolower( str )
returns a copy of
@@ -321,7 +336,7 @@
.B $0
to the next input record from the current input file;
.B getline
-.BI < file
+.BI < " file
sets
.B $0
to the next record from
@@ -359,7 +374,7 @@
in a pattern apply to the entire line.
Regular expressions may also occur in
relational expressions, using the operators
-.BR ~
+.B ~
and
.BR !~ .
.BI / re /
@@ -383,8 +398,12 @@
.br
.BI ( expr , expr,... ") in " array-name
.PP
-where a relop is any of the six relational operators in C,
-and a matchop is either
+where a
+.I relop
+is any of the six relational operators in C,
+and a
+.I matchop
+is either
.B ~
(matches)
or
@@ -405,57 +424,68 @@
and
.B END
do not combine with other patterns.
+They may appear multiple times in a program and execute
+in the order they are read by
+.IR awk .
.PP
Variable names with special meanings:
.TF FILENAME
.TP
+.B ARGC
+argument count, assignable.
+.TP
+.B ARGV
+argument array, assignable;
+non-null members are taken as filenames.
+.TP
.B CONVFMT
conversion format used when converting numbers
(default
-.BR "%.6g" )
+.BR "%.6g" ).
+.TP
+.B ENVIRON
+array of environment variables; subscripts are names.
+.TP
+.B FILENAME
+the name of the current input file.
+.TP
+.B FNR
+ordinal number of the current record in the current file.
.TP
.B FS
regular expression used to separate fields; also settable
by option
-.BI \-F fs.
+.BI \-F fs\fR.
.TP
.BR NF
-number of fields in the current record
+number of fields in the current record.
.TP
.B NR
-ordinal number of the current record
-.TP
-.B FNR
-ordinal number of the current record in the current file
-.TP
-.B FILENAME
-the name of the current input file
-.TP
-.B RS
-input record separator (default newline)
-.TP
-.B OFS
-output field separator (default blank)
-.TP
-.B ORS
-output record separator (default newline)
+ordinal number of the current record.
.TP
.B OFMT
output format for numbers (default
-.BR "%.6g" )
+.BR "%.6g" ).
+.TP
+.B OFS
+output field separator (default space).
+.TP
+.B ORS
+output record separator (default newline).
+.TP
+.B RLENGTH
+the length of a string matched by
+.BR match .
+.TP
+.B RS
+input record separator (default newline).
+.TP
+.B RSTART
+the start position of a string matched by
+.BR match .
.TP
.B SUBSEP
-separates multiple subscripts (default 034)
-.TP
-.B ARGC
-argument count, assignable
-.TP
-.B ARGV
-argument array, assignable;
-non-null members are taken as filenames
-.TP
-.B ENVIRON
-array of environment variables; subscripts are names.
+separates multiple subscripts (default 034).
.PD
.PP
Functions may be defined (at the position of a pattern-action statement) thus:
@@ -486,7 +516,7 @@
.EE
.ns
.IP
-Same, with input fields separated by comma and/or blanks and tabs.
+Same, with input fields separated by comma and/or spaces and tabs.
.PP
.EX
.nf
@@ -512,13 +542,13 @@
.fi
.EE
.SH SEE ALSO
+.IR grep (1),
.IR lex (1),
.IR sed (1)
.br
A. V. Aho, B. W. Kernighan, P. J. Weinberger,
-.I
-The AWK Programming Language,
-Addison-Wesley, 1988. ISBN 0-201-07981-X
+.IR "The AWK Programming Language" ,
+Addison-Wesley, 1988. ISBN 0-201-07981-X.
.SH BUGS
There are no explicit conversions between numbers and strings.
To force an expression to be treated as a number add 0 to it;
@@ -527,3 +557,7 @@
.br
The scope rules for variables in functions are a botch;
the syntax is worse.
+.br
+POSIX-standard interval expressions in regular expressions are not supported.
+.br
+Only eight-bit characters sets are handled correctly.
diff --git a/awk.h b/awk.h
index a36cdb1..70097b9 100644
--- a/awk.h
+++ b/awk.h
@@ -81,7 +81,8 @@
char *nval; /* name, for variables only */
char *sval; /* string value */
Awkfloat fval; /* value as number */
- int tval; /* type info: STR|NUM|ARR|FCN|FLD|CON|DONTFREE */
+ int tval; /* type info: STR|NUM|ARR|FCN|FLD|CON|DONTFREE|CONVC|CONVO */
+ char *fmt; /* CONVFMT/OFMT value used to convert from number */
struct Cell *cnext; /* ptr to next if chained */
} Cell;
@@ -109,6 +110,8 @@
#define FCN 040 /* this is a function name */
#define FLD 0100 /* this is a field $1, $2, ... */
#define REC 0200 /* this is $0 */
+#define CONVC 0400 /* string was converted from number via CONVFMT */
+#define CONVO 01000 /* string was converted from number via OFMT */
/* function types */
diff --git a/awkgram.y b/awkgram.y
index 5b5c461..e4abeed 100644
--- a/awkgram.y
+++ b/awkgram.y
@@ -86,7 +86,7 @@
%left CAT
%left '+' '-'
%left '*' '/' '%'
-%left NOT UMINUS
+%left NOT UMINUS UPLUS
%right POWER
%right DECR INCR
%left INDIRECT
@@ -357,7 +357,7 @@
| term '%' term { $$ = op2(MOD, $1, $3); }
| term POWER term { $$ = op2(POWER, $1, $3); }
| '-' term %prec UMINUS { $$ = op1(UMINUS, $2); }
- | '+' term %prec UMINUS { $$ = $2; }
+ | '+' term %prec UMINUS { $$ = op1(UPLUS, $2); }
| NOT term %prec UMINUS { $$ = op1(NOT, notnull($2)); }
| BLTIN '(' ')' { $$ = op2(BLTIN, itonp($1), rectonode()); }
| BLTIN '(' patlist ')' { $$ = op2(BLTIN, itonp($1), $3); }
diff --git a/awktest.tar b/awktest.tar
index d1cff8c..959b452 100644
--- a/awktest.tar
+++ b/awktest.tar
Binary files differ
diff --git a/b.c b/b.c
index 4f78a60..89a7841 100644
--- a/b.c
+++ b/b.c
@@ -749,7 +749,7 @@
{ "alnum", 5, isalnum },
{ "alpha", 5, isalpha },
#ifndef HAS_ISBLANK
- { "blank", 5, isspace }, /* was isblank */
+ { "blank", 5, xisblank },
#else
{ "blank", 5, isblank },
#endif
diff --git a/bugs-fixed/README b/bugs-fixed/README
new file mode 100644
index 0000000..222ef68
--- /dev/null
+++ b/bugs-fixed/README
@@ -0,0 +1,25 @@
+List of bugs fixed.
+
+1. ofs-rebuild: OFS value used to rebuild the record was incorrect.
+Fixed August 19, 2014. Revised fix August 2018.
+
+2. system-status: Instead of a floating-point division by 256, use
+the wait(2) macros to create a reasonable exit status. Fixed March 12, 2016.
+
+3. space: Use provided xisblank() function instead of ispace() for
+matching [[:blank:]].
+
+4. a-format: Add POSIX standard %a and %A to supported formats. Check
+at runtime that this format is available.
+
+5. decr-NF: Decrementing NF did not change $0. This is a decades-old
+bug. There are interactions with the old and new value of OFS as well.
+Most of the fix came from the NetBSD awk.
+
+6. string-conv: String conversions of scalars were sticky. Once a
+conversion to string happened, even with OFMT, that value was used until
+a new numeric value was assigned, even if OFMT differed from CONVFMT,
+and also if CONVFMT changed.
+
+7. unary-plus: Unary plus on a string constant returned the string.
+Instead, it should convert the value to numeric and give that value.
diff --git a/bugs-fixed/a-format.awk b/bugs-fixed/a-format.awk
new file mode 100644
index 0000000..5b7929e
--- /dev/null
+++ b/bugs-fixed/a-format.awk
@@ -0,0 +1,3 @@
+BEGIN {
+ printf("%a\n", 42)
+}
diff --git a/bugs-fixed/a-format.bad b/bugs-fixed/a-format.bad
new file mode 100644
index 0000000..1281825
--- /dev/null
+++ b/bugs-fixed/a-format.bad
@@ -0,0 +1,3 @@
+nawk: weird printf conversion %a
+ source line number 2
+%a42
diff --git a/bugs-fixed/a-format.ok b/bugs-fixed/a-format.ok
new file mode 100644
index 0000000..e421e2d
--- /dev/null
+++ b/bugs-fixed/a-format.ok
@@ -0,0 +1 @@
+0x1.5p+5
diff --git a/bugs-fixed/decr-NF.awk b/bugs-fixed/decr-NF.awk
new file mode 100644
index 0000000..7474991
--- /dev/null
+++ b/bugs-fixed/decr-NF.awk
@@ -0,0 +1,11 @@
+BEGIN {
+ $0 = "a b c d e f"
+ print NF
+ OFS = ":"
+ NF--
+ print $0
+ print NF
+ NF++
+ print $0
+ print NF
+}
diff --git a/bugs-fixed/decr-NF.bad b/bugs-fixed/decr-NF.bad
new file mode 100644
index 0000000..b634e06
--- /dev/null
+++ b/bugs-fixed/decr-NF.bad
@@ -0,0 +1,5 @@
+6
+a b c d e f
+5
+a b c d e f
+6
diff --git a/bugs-fixed/decr-NF.ok b/bugs-fixed/decr-NF.ok
new file mode 100644
index 0000000..3359cf2
--- /dev/null
+++ b/bugs-fixed/decr-NF.ok
@@ -0,0 +1,5 @@
+6
+a:b:c:d:e
+5
+a:b:c:d:e:
+6
diff --git a/bugs-fixed/ofs-rebuild.awk b/bugs-fixed/ofs-rebuild.awk
new file mode 100644
index 0000000..dd27000
--- /dev/null
+++ b/bugs-fixed/ofs-rebuild.awk
@@ -0,0 +1,17 @@
+# The bug here is that nawk should use the value of OFS that
+# was current when $0 became invalid to rebuild the record.
+
+BEGIN {
+ OFS = ":"
+ $0 = "a b c d e f g"
+ $3 = "3333"
+ # Conceptually, $0 should now be "a:b:3333:d:e:f:g"
+
+ # Change OFS after (conceptually) rebuilding the record
+ OFS = "<>"
+
+ # Unmodifed nawk prints "a<>b<>3333<>d<>e<>f<>g" because
+ # it delays rebuilding $0 until it's needed, and then it uses
+ # the current value of OFS. Oops.
+ print
+}
diff --git a/bugs-fixed/ofs-rebuild.bad b/bugs-fixed/ofs-rebuild.bad
new file mode 100644
index 0000000..7570811
--- /dev/null
+++ b/bugs-fixed/ofs-rebuild.bad
@@ -0,0 +1 @@
+a<>b<>3333<>d<>e<>f<>g
diff --git a/bugs-fixed/ofs-rebuild.ok b/bugs-fixed/ofs-rebuild.ok
new file mode 100644
index 0000000..2689218
--- /dev/null
+++ b/bugs-fixed/ofs-rebuild.ok
@@ -0,0 +1 @@
+a:b:3333:d:e:f:g
diff --git a/bugs-fixed/space.awk b/bugs-fixed/space.awk
new file mode 100644
index 0000000..6aa87d2
--- /dev/null
+++ b/bugs-fixed/space.awk
@@ -0,0 +1,22 @@
+BEGIN {
+ c[" "] = "\" \""
+ c["\a"] = "\\a"
+ c["\b"] = "\\b"
+ c["\f"] = "\\f"
+ c["\n"] = "\\n"
+ c["\r"] = "\\r"
+ c["\t"] = "\\t"
+ c["\v"] = "\\v"
+
+ sort = "LC_ALL=C sort"
+
+ for (i in c)
+ printf("%s %s [[:space:]]\n", c[i],
+ i ~ /[[:space:]]/ ? "~" : "!~") | sort
+
+ for (i in c)
+ printf("%s %s [[:blank:]]\n", c[i],
+ i ~ /[[:blank:]]/ ? "~" : "!~") | sort
+
+ close(sort)
+}
diff --git a/bugs-fixed/space.bad b/bugs-fixed/space.bad
new file mode 100644
index 0000000..f92055f
--- /dev/null
+++ b/bugs-fixed/space.bad
@@ -0,0 +1,16 @@
+" " ~ [[:blank:]]
+" " ~ [[:space:]]
+\a !~ [[:blank:]]
+\a !~ [[:space:]]
+\b !~ [[:blank:]]
+\b !~ [[:space:]]
+\f ~ [[:blank:]]
+\f ~ [[:space:]]
+\n ~ [[:blank:]]
+\n ~ [[:space:]]
+\r ~ [[:blank:]]
+\r ~ [[:space:]]
+\t ~ [[:blank:]]
+\t ~ [[:space:]]
+\v ~ [[:blank:]]
+\v ~ [[:space:]]
diff --git a/bugs-fixed/space.ok b/bugs-fixed/space.ok
new file mode 100644
index 0000000..4278c5c
--- /dev/null
+++ b/bugs-fixed/space.ok
@@ -0,0 +1,16 @@
+" " ~ [[:blank:]]
+" " ~ [[:space:]]
+\a !~ [[:blank:]]
+\a !~ [[:space:]]
+\b !~ [[:blank:]]
+\b !~ [[:space:]]
+\f !~ [[:blank:]]
+\f ~ [[:space:]]
+\n !~ [[:blank:]]
+\n ~ [[:space:]]
+\r !~ [[:blank:]]
+\r ~ [[:space:]]
+\t ~ [[:blank:]]
+\t ~ [[:space:]]
+\v !~ [[:blank:]]
+\v ~ [[:space:]]
diff --git a/bugs-fixed/string-conv.awk b/bugs-fixed/string-conv.awk
new file mode 100644
index 0000000..a1f04ab
--- /dev/null
+++ b/bugs-fixed/string-conv.awk
@@ -0,0 +1,13 @@
+BEGIN {
+ OFMT = ">>%.6g<<"
+ a = 12.1234
+ print "a =", a
+ b = a ""
+ print "1 ->", b
+ CONVFMT = "%2.2f"
+ b = a ""
+ print "2 ->", b
+ CONVFMT = "%.12g"
+ b = a ""
+ print "3 ->", b
+}
diff --git a/bugs-fixed/string-conv.bad b/bugs-fixed/string-conv.bad
new file mode 100644
index 0000000..2ab95e8
--- /dev/null
+++ b/bugs-fixed/string-conv.bad
@@ -0,0 +1,4 @@
+a = >>12.1234<<
+1 -> >>12.1234<<
+2 -> >>12.1234<<
+3 -> >>12.1234<<
diff --git a/bugs-fixed/string-conv.ok b/bugs-fixed/string-conv.ok
new file mode 100644
index 0000000..7c09711
--- /dev/null
+++ b/bugs-fixed/string-conv.ok
@@ -0,0 +1,4 @@
+a = >>12.1234<<
+1 -> 12.1234
+2 -> 12.12
+3 -> 12.1234
diff --git a/bugs-fixed/system-status.awk b/bugs-fixed/system-status.awk
new file mode 100644
index 0000000..8daf563
--- /dev/null
+++ b/bugs-fixed/system-status.awk
@@ -0,0 +1,19 @@
+# Unmodified nawk prints the 16 bit exit status divided by 256, but
+# does so using floating point arithmetic, yielding strange results.
+#
+# The fix is to use the various macros defined for wait(2) and to
+# use the signal number + 256 for death by signal, or signal number + 512
+# for death by signal with core dump.
+
+BEGIN {
+ status = system("exit 42")
+ print "normal status", status
+
+ status = system("kill -HUP $$")
+ print "death by signal status", status
+
+ status = system("kill -ABRT $$")
+ print "death by signal with core dump status", status
+
+ system("rm -f core*")
+}
diff --git a/bugs-fixed/system-status.bad b/bugs-fixed/system-status.bad
new file mode 100644
index 0000000..a1317db
--- /dev/null
+++ b/bugs-fixed/system-status.bad
@@ -0,0 +1,3 @@
+normal status 42
+death by signal status 0.00390625
+death by signal with core dump status 0.523438
diff --git a/bugs-fixed/system-status.ok b/bugs-fixed/system-status.ok
new file mode 100644
index 0000000..737828f
--- /dev/null
+++ b/bugs-fixed/system-status.ok
@@ -0,0 +1,3 @@
+normal status 42
+death by signal status 257
+death by signal with core dump status 518
diff --git a/bugs-fixed/unary-plus.awk b/bugs-fixed/unary-plus.awk
new file mode 100644
index 0000000..ba6185b
--- /dev/null
+++ b/bugs-fixed/unary-plus.awk
@@ -0,0 +1,4 @@
+BEGIN {
+ print +"q"
+ print +"43.12345678912345678"
+}
diff --git a/bugs-fixed/unary-plus.bad b/bugs-fixed/unary-plus.bad
new file mode 100644
index 0000000..76f57d5
--- /dev/null
+++ b/bugs-fixed/unary-plus.bad
@@ -0,0 +1,2 @@
+q
+43.12345678912345678
diff --git a/bugs-fixed/unary-plus.ok b/bugs-fixed/unary-plus.ok
new file mode 100644
index 0000000..90f97af
--- /dev/null
+++ b/bugs-fixed/unary-plus.ok
@@ -0,0 +1,2 @@
+0
+43.1235
diff --git a/lib.c b/lib.c
index 1b6d86e..ba6ebd4 100644
--- a/lib.c
+++ b/lib.c
@@ -356,6 +356,7 @@
}
}
setfval(nfloc, (Awkfloat) lastfld);
+ donerec = 1; /* restore */
if (dbg) {
for (j = 0; j <= lastfld; j++) {
p = fldtab[j];
@@ -387,6 +388,19 @@
setfval(nfloc, (Awkfloat) n);
}
+void setlastfld(int n) /* set lastfld cleaning fldtab cells if necessary */
+{
+ if (n > nfields)
+ growfldtab(n);
+
+ if (lastfld < n)
+ cleanfld(lastfld+1, n);
+ else
+ cleanfld(n+1, lastfld);
+
+ lastfld = n;
+}
+
Cell *fieldadr(int n) /* get nth field */
{
if (n < 0)
diff --git a/main.c b/main.c
index 45b4dbb..d292ad3 100644
--- a/main.c
+++ b/main.c
@@ -22,7 +22,7 @@
THIS SOFTWARE.
****************************************************************/
-const char *version = "version 20130105";
+const char *version = "version 20130105 + fixes by arnold@skeeve.com 20180822";
#define DEBUG
#include <stdio.h>
diff --git a/makefile b/makefile
index ff54372..8f5906f 100644
--- a/makefile
+++ b/makefile
@@ -26,13 +26,13 @@
CFLAGS = -O2
CFLAGS =
-CC = gcc -Wall -g -Wwrite-strings
+#CC = gcc -Wall -g -Wwrite-strings
CC = gcc -g -Wall -pedantic
-CC = gcc -O4 -Wall -pedantic -fno-strict-aliasing
-CC = gcc -fprofile-arcs -ftest-coverage # then gcov f1.c; cat f1.c.gcov
+#CC = gcc -O4 -Wall -pedantic -fno-strict-aliasing
+#CC = gcc -fprofile-arcs -ftest-coverage # then gcov f1.c; cat f1.c.gcov
YACC = bison -d -y
-YACC = yacc -d
+# YACC = yacc -d
#YFLAGS = -d -S
# -S uses sprintf in yacc parser instead of sprint
diff --git a/maketab.c b/maketab.c
index 31acd75..e23974c 100644
--- a/maketab.c
+++ b/maketab.c
@@ -62,6 +62,7 @@
{ DIVIDE, "arith", " / " },
{ MOD, "arith", " % " },
{ UMINUS, "arith", " -" },
+ { UPLUS, "arith", " +" },
{ POWER, "arith", " **" },
{ PREINCR, "incrdecr", "++" },
{ POSTINCR, "incrdecr", "++" },
diff --git a/proto.h b/proto.h
index 9a657ef..ad6f2e8 100644
--- a/proto.h
+++ b/proto.h
@@ -124,6 +124,7 @@
extern void fldbld(void);
extern void cleanfld(int, int);
extern void newfld(int);
+extern void setlastfld(int);
extern int refldbld(const char *, const char *);
extern void recbld(void);
extern Cell *fieldadr(int);
@@ -193,3 +194,5 @@
extern FILE *popen(const char *, const char *);
extern int pclose(FILE *);
+
+extern const char *flags2str(int flags);
diff --git a/run.c b/run.c
index d3f9c68..c818a08 100644
--- a/run.c
+++ b/run.c
@@ -31,6 +31,8 @@
#include <string.h>
#include <stdlib.h>
#include <time.h>
+#include <sys/types.h>
+#include <sys/wait.h>
#include "awk.h"
#include "ytab.h"
@@ -323,14 +325,18 @@
{
Cell *y;
+ /* copy is not constant or field */
+
y = gettemp();
+ y->tval = x->tval & ~(CON|FLD|REC);
y->csub = CCOPY; /* prevents freeing until call is over */
y->nval = x->nval; /* BUG? */
- if (isstr(x))
+ if (isstr(x) /* || x->ctype == OCELL */) {
y->sval = tostring(x->sval);
+ y->tval &= ~DONTFREE;
+ } else
+ y->tval |= DONTFREE;
y->fval = x->fval;
- y->tval = x->tval & ~(CON|FLD|REC|DONTFREE); /* copy is not constant or field */
- /* is DONTFREE right? */
return y;
}
@@ -817,6 +823,17 @@
char *buf = *pbuf;
int bufsize = *pbufsize;
+ static int first = 1;
+ static int have_a_format = 0;
+
+ if (first) {
+ char buf[100];
+
+ sprintf(buf, "%a", 42.0);
+ have_a_format = (strcmp(buf, "0x1.5p+5") == 0);
+ first = 0;
+ }
+
os = s;
p = buf;
if ((fmt = (char *) malloc(fmtsz)) == NULL)
@@ -859,6 +876,12 @@
adjbuf(&buf, &bufsize, fmtwd+1+p-buf, recsize, &p, "format4");
switch (*s) {
+ case 'a': case 'A':
+ if (have_a_format)
+ flag = *s;
+ else
+ flag = 'f';
+ break;
case 'f': case 'e': case 'g': case 'E': case 'G':
flag = 'f';
break;
@@ -901,6 +924,8 @@
p += strlen(p);
sprintf(p, "%s", t);
break;
+ case 'a':
+ case 'A':
case 'f': sprintf(p, fmt, getfval(x)); break;
case 'd': sprintf(p, fmt, (long) getfval(x)); break;
case 'u': sprintf(p, fmt, (int) getfval(x)); break;
@@ -1003,7 +1028,7 @@
x = execute(a[0]);
i = getfval(x);
tempfree(x);
- if (n != UMINUS) {
+ if (n != UMINUS && n != UPLUS) {
y = execute(a[1]);
j = getfval(y);
tempfree(y);
@@ -1033,6 +1058,8 @@
case UMINUS:
i = -i;
break;
+ case UPLUS: /* handled by getfval(), above */
+ break;
case POWER:
if (j >= 0 && modf(j, &v) == 0.0) /* pos integer exponent */
i = ipow(i, (int) j);
@@ -1479,6 +1506,7 @@
Node *nextarg;
FILE *fp;
void flush_all(void);
+ int status = 0;
t = ptoi(a[0]);
x = execute(a[1]);
@@ -1515,7 +1543,20 @@
break;
case FSYSTEM:
fflush(stdout); /* in case something is buffered already */
- u = (Awkfloat) system(getsval(x)) / 256; /* 256 is unix-dep */
+ status = system(getsval(x));
+ u = status;
+ if (status != -1) {
+ if (WIFEXITED(status)) {
+ u = WEXITSTATUS(status);
+ } else if (WIFSIGNALED(status)) {
+ u = WTERMSIG(status) + 256;
+#ifdef WCOREDUMP
+ if (WCOREDUMP(status))
+ u += 256;
+#endif
+ } else /* something else?!? */
+ u = 0;
+ }
break;
case FRAND:
/* in principle, rand() returns something in 0..RAND_MAX */
diff --git a/tran.c b/tran.c
index a9fa325..06f32fc 100644
--- a/tran.c
+++ b/tran.c
@@ -67,6 +67,18 @@
extern Cell **fldtab;
+static void
+setfree(Cell *vp)
+{
+ if (&vp->sval == FS || &vp->sval == RS ||
+ &vp->sval == OFS || &vp->sval == ORS ||
+ &vp->sval == OFMT || &vp->sval == CONVFMT ||
+ &vp->sval == FILENAME || &vp->sval == SUBSEP)
+ vp->tval |= DONTFREE;
+ else
+ vp->tval &= ~DONTFREE;
+}
+
void syminit(void) /* initialize symbol table with builtin vars */
{
literal0 = setsymtab("0", "0", 0.0, NUM|STR|CON|DONTFREE, symtab);
@@ -282,6 +294,7 @@
{
int fldno;
+ f += 0.0; /* normalise negative zero to positive zero */
if ((vp->tval & (NUM | STR)) == 0)
funnyvar(vp, "assign to");
if (isfld(vp)) {
@@ -290,13 +303,18 @@
if (fldno > *NF)
newfld(fldno);
dprintf( ("setting field %d to %g\n", fldno, f) );
+ } else if (&vp->fval == NF) {
+ donerec = 0; /* mark $0 invalid */
+ setlastfld(f);
+ dprintf( ("setting NF to %g\n", f) );
} else if (isrec(vp)) {
donefld = 0; /* mark $1... invalid */
donerec = 1;
}
if (freeable(vp))
xfree(vp->sval); /* free any previous string */
- vp->tval &= ~STR; /* mark string invalid */
+ vp->tval &= ~(STR|CONVC|CONVO); /* mark string invalid */
+ vp->fmt = NULL;
vp->tval |= NUM; /* mark number ok */
if (f == -0) /* who would have thought this possible? */
f = 0;
@@ -318,6 +336,7 @@
{
char *t;
int fldno;
+ Awkfloat f;
dprintf( ("starting setsval %p: %s = \"%s\", t=%o, r,f=%d,%d\n",
(void*)vp, NN(vp->nval), s, vp->tval, donerec, donefld) );
@@ -332,16 +351,28 @@
} else if (isrec(vp)) {
donefld = 0; /* mark $1... invalid */
donerec = 1;
+ } else if (&vp->sval == OFS) {
+ if (donerec == 0)
+ recbld();
}
- t = tostring(s); /* in case it's self-assign */
+ t = s ? tostring(s) : tostring(""); /* in case it's self-assign */
if (freeable(vp))
xfree(vp->sval);
- vp->tval &= ~NUM;
+ vp->tval &= ~(NUM|CONVC|CONVO);
vp->tval |= STR;
- vp->tval &= ~DONTFREE;
+ vp->fmt = NULL;
+ setfree(vp);
dprintf( ("setsval %p: %s = \"%s (%p) \", t=%o r,f=%d,%d\n",
(void*)vp, NN(vp->nval), t,t, vp->tval, donerec, donefld) );
- return(vp->sval = t);
+ vp->sval = t;
+ if (&vp->fval == NF) {
+ donerec = 0; /* mark $0 invalid */
+ f = getfval(vp);
+ setlastfld(f);
+ dprintf( ("setting NF to %g\n", f) );
+ }
+
+ return(vp->sval);
}
Awkfloat getfval(Cell *vp) /* get float val of a Cell */
@@ -373,17 +404,78 @@
fldbld();
else if (isrec(vp) && donerec == 0)
recbld();
- if (isstr(vp) == 0) {
- if (freeable(vp))
- xfree(vp->sval);
- if (modf(vp->fval, &dtemp) == 0) /* it's integral */
- sprintf(s, "%.30g", vp->fval);
- else
- sprintf(s, *fmt, vp->fval);
- vp->sval = tostring(s);
- vp->tval &= ~DONTFREE;
- vp->tval |= STR;
+
+ /*
+ * ADR: This is complicated and more fragile than is desirable.
+ * Retrieving a string value for a number associates the string
+ * value with the scalar. Previously, the string value was
+ * sticky, meaning if converted via OFMT that became the value
+ * (even though POSIX wants it to be via CONVFMT). Or if CONVFMT
+ * changed after a string value was retrieved, the original value
+ * was maintained and used. Also not per POSIX.
+ *
+ * We work around this design by adding two additional flags,
+ * CONVC and CONVO, indicating how the string value was
+ * obtained (via CONVFMT or OFMT) and _also_ maintaining a copy
+ * of the pointer to the xFMT format string used for the
+ * conversion. This pointer is only read, **never** dereferenced.
+ * The next time we do a conversion, if it's coming from the same
+ * xFMT as last time, and the pointer value is different, we
+ * know that the xFMT format string changed, and we need to
+ * redo the conversion. If it's the same, we don't have to.
+ *
+ * There are also several cases where we don't do a conversion,
+ * such as for a field (see the checks below).
+ */
+
+ /* Don't duplicate the code for actually updating the value */
+#define update_str_val(vp) \
+ { \
+ if (freeable(vp)) \
+ xfree(vp->sval); \
+ if (modf(vp->fval, &dtemp) == 0) /* it's integral */ \
+ sprintf(s, "%.30g", vp->fval); \
+ else \
+ sprintf(s, *fmt, vp->fval); \
+ vp->sval = tostring(s); \
+ vp->tval &= ~DONTFREE; \
+ vp->tval |= STR; \
}
+
+ if (isstr(vp) == 0) {
+ update_str_val(vp);
+ if (fmt == OFMT) {
+ vp->tval &= ~CONVC;
+ vp->tval |= CONVO;
+ } else {
+ /* CONVFMT */
+ vp->tval &= ~CONVO;
+ vp->tval |= CONVC;
+ }
+ vp->fmt = *fmt;
+ } else if ((vp->tval & DONTFREE) != 0 || ! isnum(vp) || isfld(vp)) {
+ goto done;
+ } else if (isstr(vp)) {
+ if (fmt == OFMT) {
+ if ((vp->tval & CONVC) != 0
+ || ((vp->tval & CONVO) != 0 && vp->fmt != *fmt)) {
+ update_str_val(vp);
+ vp->tval &= ~CONVC;
+ vp->tval |= CONVO;
+ vp->fmt = *fmt;
+ }
+ } else {
+ /* CONVFMT */
+ if ((vp->tval & CONVO) != 0
+ || ((vp->tval & CONVC) != 0 && vp->fmt != *fmt)) {
+ update_str_val(vp);
+ vp->tval &= ~CONVO;
+ vp->tval |= CONVC;
+ vp->fmt = *fmt;
+ }
+ }
+ }
+done:
dprintf( ("getsval %p: %s = \"%s (%p)\", t=%o\n",
(void*)vp, NN(vp->nval), vp->sval, vp->sval, vp->tval) );
return(vp->sval);
@@ -457,3 +549,37 @@
*bp++ = 0;
return (char *) buf;
}
+
+const char *flags2str(int flags)
+{
+ static const struct ftab {
+ const char *name;
+ int value;
+ } flagtab[] = {
+ { "NUM", NUM },
+ { "STR", STR },
+ { "DONTFREE", DONTFREE },
+ { "CON", CON },
+ { "ARR", ARR },
+ { "FCN", FCN },
+ { "FLD", FLD },
+ { "REC", REC },
+ { "CONVC", CONVC },
+ { "CONVO", CONVO },
+ { NULL, 0 }
+ };
+ static char buf[100];
+ int i;
+ char *cp = buf;
+
+ for (i = 0; flagtab[i].name != NULL; i++) {
+ if ((flags & flagtab[i].value) != 0) {
+ if (cp > buf)
+ *cp++ = '|';
+ strcpy(cp, flagtab[i].name);
+ cp += strlen(cp);
+ }
+ }
+
+ return buf;
+}