BUGS

    Is it fixed?

    Fixed implies a permenent condition which I cannot guarantee captain,
    I would rather say that it is well patched.  	    

					     -- Commander Data



131. pivottable may sort forever when asked for the median	Patched 3.4.5

130. Invalid FITS header in talble to fits			Patched	3.4.5

129. updatetable typo'ed					Patched 3.3.12

128. pivottable fails if a column is named "N"			Patched 3.3.12

127. column formatting - buffer not initialized.		Patched 3.3.11

    The following bug appeared in "column" a while back, and I am
    finally reporting it.  If I take a dummy input file
	x	y	z
	-	-	-
	1	100	200
	2
	3	100	200

    and then format the columns with x%.1f y%.1f z%.1f I get

	x       y	z
	-       -	-
	1.0     100.0	200.0
	2.0     2.	2.
	3.0     100.0	200.0

    Notice how the empty entries in row 2 have been erroneously filled in with
    the previous entry (columns y and z are for some reason set to the value
    of column x).

    What used to happen is that the empty entries would be set to 0.0
    for %1.f, although it might make as much sense to be left empty.



126. If a table with a column named "e" is used with 		Patched 3.3.11
     "row" or one of its links and the input selection
     expression contains a literal number in scientific
     notation (ie 10e-1), the "e" in the exponential notation
     is replaced with the column value causing very difficult
     to find errors.

125. The new add column feature in table invoked by +COL	Patched 3.3.11
     was getting tripped by ++var

124. When justifying more than 64000 lines  			Patched 3.3.11
     the temp file is handled incorrectly and
     only 64000 lines of output are produced.

	#77 from below resurfaced!

123. statstable - Added -l option to label stats when called	Patched 3.3.11
     as one of the sub stat links.

122. tabletofits - Uninitialized local causes segv.		Patched 3.3.11 

121. statstable, mediantable - statistical value were		Patched 3.3.11
     only reported to the precision of the most precise
     value in the input column.  The median of odd numbers
     of rows was computed incorrectly.
     
120. When sorting (indexing) a pair of columns with		Patched 3.3.10
     a modulus modifier the two bins adjacent to zero
     would be sorted in the same bin.

119. Column and friends - When comparing column names 		Patched 3.3.10
     given on the command line to those in the table 
     leading and trailing '_' characters were ignored
     along with spaces.

	This was obviously coded in purposfully but
	is a bug.

118. table_loadva still contained 32 bit file offsets!		Patched 3.3.10

117. The API to table_loadva was used inconsistantly in		Patched 3.3.10
     internal calls.  When fixed up two arguments were 
     swapped.
    

116. Tawk replaces @ with the _Stack even when called 		Patched 3.3.9
     wo/stack enabled.

115. Tawk truncates seconds in conversion of mjd to unix	Patched 3.3.9
     seconds.

114. Release to Eric, documenting the fix for sorttable -q	Patched 3.3.8

113. TABLEFUNCTIONS cannot be nulled out by a script.		Patched 3.3.7
     The compiled in default is always searched.

	This issue effected radebox, radexcirc and all of the 
	units and coordinate system scripts.
	
112. radecbox never checks to see if it is actaully 		Patched 3.3.7
     called as radeccirc.
 
111. edittable is documented as reporting the number of		Patched 3.3.7
     rows edited when the -v switch is given.  The code was 
     there but the counter was never incremented.

110. column rename was damaged when the pattern matching	Patched 3.3.7
     code was added.

109. search code to find records across the wrap point		Patched 3.3.3
     in a paired indexed table was hopelessly typoed.

     More of #101 below.

108. jointable - when the join column is the last column of	Patched 3.3.3
	table 1 and table 2 contains only one column the
	field separators are not output in the header.

	This is similar to #68 and #60.  Added tests.

107. column - when adding a new column the forced width 	Unpatched
	formatting ",##" doesn't work.

Release 3.3.2

106. search - dual indexed columns with negative value  	Patched 3.3.2
	When searching in a paired indexed table with negative
	range values the modulus expression on the lower bin
	bound would round towards zero, this would cause the
	lowest bin to be skipped. (#98 below)

	This was still broken for the bin just below 0!

105. header eats up the table by accident!			Patched 3.3.2

     This breaks headtable, tailtable and others

104. search will propigate corrupt data without warning		Patched 3.2 11
     when the input key table has to few/to many columns in
     a row.  This can shift all columns in joined data
     output.

103. table - row, compute and friends				Patched 3.2.11
     When a double quote (") was encountered immedietly
     following a column name the quote character was
     missed.  This caused any following column name to
     be missed.

     This may fix bug #65 reported below.

102. search segv on short tables while computing averge		Patched 3.2.10
     row length in bytes.

101. search code to find records across the wrap point		Patched 3.2.9
     in a paired indexed table was hopelessly typoed.

100. Some starbase scripts change PATH to include		Patched 3.2.9
     the release directory but do not explicitly
     re-export the new PATH value.  This fails 
     under solaris /bin/sh.

	This is the same bug as #76 below

 99. search - a remaining long long file offset			Patched 3.2.9
	variable was declaired as int.  Caused
	search to fail of very large files.

 98. search - dual indexed columns with negative value  	Patched 3.2.9
	When searching in a paired indexed table with negative
	range values the modulus expression on the lower bin
	bound would round towards zero, this would cause the
	lowest bin to be skipped. 

 97. table, compute and friends					Patched 3.2.9

	The sign of number of numbers between 0 and -1 is dropped
	when converting to sexagasmal format (0:00:00). This
	was reintrobuced when fixing #93.  Cases are now in
	the test suite.

 96. search trims the input key table to force 			Patched 3.2.9
     proper trimming of the key value in compare.
     This solution was too drastic and created
     non-obvious result tables when the key table
     and the search table are the same and -j is
     used.

     	This appears to be the 4th time I've fixed
	this (#83, #43, #30).

 95. mjd refuses to convert dates with				Patched 3.2.9
     seconds value > 59 (should be > 60)

 94. Timezone is incorrectly defaulted to GMT on the 		Patched 3.2.9
     first call to unx() and the local zone on
     subsequent calls.

 93. Disconnected minus sign.					Patched 3.2.9
     Same as #66 below.  Fix implimented there was insufficient

     Added tests in test/tawk/saostrtod

 92. search -- -C[123] and -S[123] options ignore		Pathced 3.2.9
     numeric indicies created with index -n but 
     see indicees created with index -h

 91. search -- an uninitialized value in the axis		Pathced 3.2.9
     structure caused a failure of the -C[123] option
     under linux.

 90. fitsbase -- "\f" is not interpolated to ^L by		Pathced 3.2.9
     default under linux,

 89. index -- uninitialized variable in zones code		Pathced 3.2.5
     causes random segv.

 88. search -S2 and friends should warn users if		Patched 3.2.5
     index files are not numeric

 87. search -S2 and friends no longer work if there		Patched 3.2.5
     are no index files.

 86. row, compute and friends segv on quoted strings 		Patched 3.2.3
     longer than 400 bytes.  Increased limit to 8192.
     The larger limit was lost when porting to mawk 1.3.3

	This is the same bug as #40 below.

 85. svdfittable outputs garbage rows results when fed		Patched 3.2.3
     tables with rows of differing numbers of columns

 84. column segv with -hV -t and the second of two		Patched 3.2.2
     headers is the only output.

 83. search fails when input search keys contain 		Patched 3.2.2
     trailing spaces.

     	This appears to be the 3rd time I've fixed this (#43, #30).

 82. sorttable crashed on linux for files > 1Meg		Patched 3.2.2

 	linux setbuffer (glibc) brain damaged.

 81. Only row and column based filters recognize		Fix not planned
     formfeed delimited tables.  Only column accepts
     the -t option.
			
 80. sorttable -c can report the wrong line number 		Patched 3.1.8
     when it finds disorder.

 79. searching with a table of one record crashes		Patched 3.1.8

 78. median in statstable fails for tables with 1 row		Patched 3.1.8

 77. When justifying more than 64000 lines from a 		Patched 3.1.8
     pipe the temp file is handled incorrectly and
     only 64000 lines of output are produced.
     
 76. PATH set but not exported in shell scripted		Patched 3.1.8
     commands

 75. report doesn't output a \ when it sees \\			Patched 3.1.8

 74. slalink - eqgal eg1950					Patched 3.1.7

 	functions return input coords instead of results 
 	
 73. table, compute and friends					Patched 3.1.7

	The sign of number of numbers between 0 and -1 is dropped
	when converting to sexagasmal format (0:00:00).

 73. table, row and friends ...					Fix not planned

 	Cannot use column names in reqular expressions.  They
	are incorrectly substituted with column numbers.

 72. statstable, meantable ...					Patched 3.1.3

 	-w switch broken

 71. search segv.						Patched 3.1.2
 
 	STRICTTAB not the default when calling rowget wo/ a header.

 70. column segv. when deleteing and adding columns.		Patched 3.1.2

      column < infile -a Num:x DATE_TIME:x DDelta:x slope delta AdjCnt C1 C2 T1 T2 Err
      can crash column.

	Column selection code contained a typo, using the wrong index when 
	filling in the selected column numbers.

 69. table							Patched 3.1.2

 	segv possible if any included file is of zero size.

 68. jointable							Patched 3.1.2

	When table 2 contains only the join column
	an extra field sepatator is output in the header.


 67. sorttable -i doesn't work at all.				Patched 3.1.1

 66. Disconnected minus sign.					Patched 3.1.1

    From bmcleod@angus.harvard.edu  Fri Jun  1 21:04:12 2001
    To: john@angus.harvard.edu
    Subject: table formatting problem -- disconnected minus sign

    bmcleod angus 108> printf "X\n-\n-1\n" | table -h '{printf "%5.2@", X}'
    -    1:00:00.00bmcleod angus 109> 


 65. Extreme strange behavoir in column replacement		Patched 3.2.11

	This appears to be a confusing manifestation of
	bug #103 above.

     # Extreme strange behavoir
     #
     table -D -h '{ print '"$1"'"="'" $2 "' }'

     The lexer does the wrong thing here?


 64. jointable would mangle the data in file2 after		Patched 2.7.8
     reaching and eof on file1.  Only noticed with 
     -a2 option.

 63. tawk would convert the string "5ddd" to the		Patched 2.7.8
     number 5 instead of 0. (Unpatched later).


 62. fk54z fk45z						Patched 2.7.0

	bepoch parameter to low level slalib routines is a
	modern epoch value not an mjd.

 61. fixtotable							Patched 2.6.11

	Segv with more than 4 extracted coumns.

 60. jointable							Patched 2.6.9

	When table 2 contains only the join column
	no record sepatators are output in the header.

 59. statstable refused to compute for columns which		Patched 2.6.9
     contained scientific notation.

 58. medtable was a copy of rmstable, selecting the		Patched 2.6.9
     RMS row instead of the Median row.

 57. column mixes up the output columns when asked		Patched 2.6.8
     select and format a table in the same invokation

 56. search without -j leaked the results union on 		Patched 2.6.8
     each keytable row.  For largish inputs this would
     slow the program to a crawl.

 55. pastetable could segv if the number of rows		Patched 2.6.8
     in the input tables were different by more
     than one.

 54. listtotable did not trim leading and trailing		Patched 2.6.8
     spaces from the field names and generated
     ugly dashlines.

 53. pastetable failed to output header values if they 		Patched 2.6.8
     are available in the first file to paste.

 52. In column the absolute width of added columns 
     cannot be set with the -w switch.

 51. obscure bug in search can cause segv.			Patched 2.6.8

 50. statstable min and range.					Patched 2.6.7

	min is computed incorrectly for negative numbers.

 49. column strikeout						Patched 2.6.7

	Striking out columns with -x can cause a segv.

 48. Free unallocated column format vectors.			Patched 2.6.6

	This was causing plottable and column to trap.

 47. Fix typo in sorttable that made it unusable		Patched 2.6.5


--- Release 2.6.0 Here

 46. The mjd function would not recognize mjd numbers.		Patched 2.5

 45. Proper expansion of the TABLEFUNCTIONS path.		Patched 2.5

	The util routine FileList only expaneded a single directory.
	It was supposed to be a path of directories.

 44. Cannot justify a number to the left or center		Patched 2.5
     and cannot force text justification of a numeric column.

 43. search was not skipping over leading spaces.		Patched 2.5

     This resulted in a mismatch in the search and sort colating sequences.
	
 42. cal function returns ? as the day value for		Patched 2.5
     dates before 1970.

	Since cal is based on strftime it seems that dates before the UNIX
	epoch are too confusing to get the day value right.  The default 
	date format has been changes so that is doesn't include the day 
	of the week.

 41. diffrtable uses poor code to create a dashline.   		 Patched 2.5

	intertable suffered from the same problem

 40. tawk fixed length for quoted strings encountered.		Patched 2.5

	There is a fixed length buffer for parsing quoted strings
	in the mawk compiler.  This value was 400 now 4096.

 39. sorttable [-u -U -d -D] and large inputs.			Patched 2.5

	For large input files, (>100,000 lines) and unique line output
	produces incorrect output.

	Fixes for Bug #27 broke several other things.


 38. fk54 -- incorrect calling sequence to slalib		Patched 2.5


 37. floating point overflow on tawk

	Text in a column that "looks" like a number but is actually just
    	text is converted anyway and can cause floating point overflows.


 36. mdj() tawk builtin function.				Patched 2.5

	mjd returns values 365 days to large for days in the month of
	febuary.

	Bug in slacldj fixed in newer releases of sla.

 35. cal() tawk builtin function.				Patched 2.5

	calling cal() crashes the program.


 34. sort -n and -h						Patched 2.5

	"+" in front of a number is treated as TEXT.


 33. radecbox, radeccirc					Patched 2.5

	Negative declinations are not included in searches.

	The declination axis is incorrectly identified as a circular coordinate
	system to search.  Negative values are "wrapped" around 360 degrees.


 32. fixtotable	- table_colarg();				Patched 2.5

	Using a column flag more than once causes a segv while printing
	the error message!

 31. index							Patched 2.5

	index -mi with invalid table input SEGV.

 30. search							Patched 2.5

	Search should trim fields of leading and training whitespace before 
	doing the compare.

 29. row 							Patched 2.3d

	Row does not concatenate its arguments into a program as the 
	/rdb version does.

 28. search on HP-UX						Patched 2.3d

	search fails on HP-UX returning too few rows.  This was an obvious
	codding error in the sorting compare function in range.c.  I don't
	know why search seemed to work on other machines (different qsort
	implimentations?).

 27. sorttable [-u -U -d] and large inputs.			Patched 2.3d

	For large input files, (>100,000 lines) and unique line output
	_sort dumps core.

	The saved line buffer was being tested before initialised on the
	first iteration.  

	For large inputs the -U and -d options produce incorrect results.

	The uniqueness of a line was being evaluated on a buffer by buffer
	basis instead of over the entire sort.

 26. radecbox and radeccirc					Patched 2.3d

	negative declinations are not searched.

	obvious coding error in the parameter range checking call.

	stupid code -n table coppies all "option" arguments to the front of 
	the argument list before calling tawk.  This caused the negative
	dec to be interpreted as the tawk program, a single number evaluation.

 25. jointable							Patched 2.3d

	If the join column in the two tables were not in the same
	column number the join fails

 24. jointable							Patched 2.3d

	When the last column of a table being joined ends in a space an 
	extra column was inserted into the output.

 23. table_colarg routine					Patched 2.3d

	String column values are not NULL terminated on return.

 22. search							Patched 2.3d

	Greater than and less than comparisons are reversed for string column.

 21. sorttable and jointable call sort and join			Patched 2.3d

	Installing modified versions of familiar system programs was a bad
	idea.  The starbase versions of sort and join have been renamed to
	_sort and _join.

 20. fldtotable							Patched 2.3d

	When the field number begins with 9 there are many extra spaces output
	in the database column.  The field start and stop columns where not
	being cast from strings to numbers.  Incrementing a string "9" is a
	string "10" which is still less than the number 9.  The string would
	continue to increment until "90".

	Inputs are now explicitly cast to numbers.

 19. jointable							Patched 2.5

	When two table with a column of the same name are joined (not the
	join column) the columns are not renamed Col_1 and Col_2 as they 
	are in the pastetable, cprodtable and search programs.

 18. table programs						Patched 2.3b

	Header value identifiers containing "_" characters were not
	included as values in the generated awk program.

 17. jointable							Patched 2.3b

	jointable sometimes corupts the last field in a file.

	The field "trim" feature was testing uninitialized data.

 16. sorttable -h.						Patched	2.3b

	sorttable fails when sorting on sexagesmal values in more than one 
	column.

	Code for the "optimized" first field case was accidently coppied
	into the normal sexagasmal compare branch.  When more than one
	field of sexagasmal is compared, the cached first field value
	is corupted.

 15. Table.  header array value initializarion.			Patched 2.3b

	A newline was inserted in the awk program after the header
	value instead of a comma.

	Introduced during the addition of column array feature.
	Released because of a lack of testing.

 14. Table.  uninitialized pointer tfp.				Patched 2.2

 13. Search -- alpha						Patched 2.3d

	even under the DEC compiler there is a bug in search.

	The code does not seem to be very 64 bit friendly.  This bug
	appears to be the same as #28 above.

 12. Complete melt down of all SAO written code on the alpha!

	This was a result of a bug in gcc-2.7.0.  Declairation for malloc
	that are not in actual source file are ignored???

	Work around is to use the DEC compiler, which is faster anyway.

 11. listtotable/ tabletolist					Patched 2.3c

     These program where woefully under tested.  There should now be some 
     expectation that they will work as in /rdb.

 10. tawk							Patched 2.2

     In a function declairation without an explicite return statement
     the number of return values goes uninitialized and the program 
     issues a warning.

     Initialize the nrets member of the active_funct structure with 0.

  9. Missing Documentation:

	These are all programs added in this release (2.24).

	validate						Patched 2.5
	intertable						Patched 2.5
	diffrtable						Patched 2.5
	uniontable						Patched 2.5
	tabletolist
	listtotable
	prec							Program deleted
	grestable
	tabletosed
	trantable
	tabletom4

  8. Search							Patched 2.2

    Routine table_rangerows used many times in the search program passes
    the wrong "end" range value to the function it calls.

  7. Configuration,  libdir needed for install.

  6. Configuration,  man instalation.				Patched 2.2

    man/html/help directories populated in distribution.

  5. Configuration, lex/yacc nedded to compile.			Patched 2.2

  4. Search					 		Patched 2.2

    Searching 2d spherical coordinate system.

	The routine which computes the search box multiplies by the
	cosine of the declination instead of dividing.

  3. Sort					 		Patched 2.2

    Major bug in sort.  All sorting and indexing is suspect.

	An optimization introduced in the 2.21 internal release
	caused sort not to sort at all.  The problem did not 
	make it into any released versions.

  2. Search

    Search on a hash index does not work.  Never tested when written.
    this should take a couple of hours to fixup.

  1. Column rename						Patched 2.5

    Rename broken on machines where sizeof(int) != sizeof(char *).

	poor programming practice.