summaryrefslogtreecommitdiff
path: root/TODO
blob: 2878d563502aee1885e62cea4b7151a342eb5e2f (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
restore djgpp, eventually
merge TODO lists
add unit tests for lib/*.c

strip: add an option to specify the program used to strip binaries.
  suggestion from Karl Berry

doc/coreutils.texi:
  Address this comment: FIXME: mv's behavior in this case is system-dependent
  Better still: fix the code so it's *not* system-dependent.

implement --target-directory=DIR for install (per texinfo documentation)

ls: add --format=FORMAT option that controls how each line is printed.

cp --no-preserve=X should not attempt to preserve attribute X
  reported by Andreas Schwab

copy.c: Address the FIXME-maybe comment in copy_internal.
And once that's done, add an exclusion so that `cp --link'
no longer incurs the overhead of saving src. dev/ino and dest. filename
in the hash table.

See if we can be consistent about where --verbose sends its output:
  These all send --verbose output to stdout:
    head, tail, rm, cp, mv, ln, chmod, chown, chgrp, install, ln
  These send it to stderr:
    shred mkdir split
  readlink is different

Write an autoconf test to work around build failure in HPUX's 64-bit mode.
See notes in README -- and remove them once there's a work-around.

Integrate use of sendfile, suggested here:
  http://mail.gnu.org/archive/html/bug-fileutils/2003-03/msg00030.html
I don't plan to do that, since a few tests demonstrate no significant benefit.

Should printf '\0123' print "\n3"?
  per report from TAKAI Kousuke on Mar 27
  http://mail.gnu.org/archive/html/bug-coreutils/2003-03/index.html

printf: consider adapting builtins/printf.def from bash

df: add `--total' option, suggested here http://bugs.debian.org/186007

seq: give better diagnostics for invalid formats:
   e.g. no or too many % directives
seq: consider allowing format string to contain no %-directives

m4: rename all macros that start with AC_ to start with another prefix

resolve RH report on cp -a forwarded by Tim Waugh

Martin Michlmayr's patch to provide ls with `--sort directory' option

tail: don't use xlseek; it *exits*.
  Instead, maybe use a macro and return nonzero.

add mktemp?  Suggested by Nelson Beebe

df: alignment problem of `Used' heading with e.g., -mP
  reported by Karl Berry

tr: support nontrivial equivalence classes, e.g. [=e=] with LC_COLLATE=fr_FR

lib/strftime.c: Since %N is the only format that we need but that
  glibc's strftime doesn't support, consider using a wrapper that
  would expand /%(-_)?\d*N/ to the desired string and then pass the
  resulting string to glibc's strftime.

sort: Compress temporary files when doing large external sort/merges.
  This improves performance when you can compress/uncompress faster than
  you can read/write, which is common in these days of fast CPUs.
  suggestion from Charles Randall on 2001-08-10

sort: Add an ordering option -R that causes 'sort' to sort according
  to a random permutation of the correct sort order.  Also, add an
  option --random-seed=SEED that causes 'sort' to use an arbitrary
  string SEED to select which permutations to use, in a deterministic
  manner: that is, if you sort a permutation of the same input file
  with the same --random-seed=SEED option twice, you'll get the same
  output.  The default SEED is chosen at random, and contains enough
  information to ensure that the output permutation is random.
  suggestion from Feth AREZKI, Stephan Kasal, and Paul Eggert on 2003-07-17

unexpand: [http://www.opengroup.org/onlinepubs/007908799/xcu/unexpand.html]
  printf 'x\t \t y\n'|unexpand -t 8,9 should print its input, unmodified.
  printf 'x\t \t y\n'|unexpand -t 5,8 should print "x\ty\n"

Let GNU su use the `wheel' group if appropriate.
  (there are a couple patches, already)

sort: Investigate better sorting algorithms; see Knuth vol. 3.

  We tried list merge sort, but it was about 50% slower than the
  recursive algorithm currently used by sortlines, and it used more
  comparisons.  We're not sure why this was, as the theory suggests it
  should do fewer comparisons, so perhaps this should be revisited.
  List merge sort was implemented in the style of Knuth algorithm
  5.2.4L, with the optimization suggested by exercise 5.2.4-22.  The
  test case was 140,213,394 bytes, 426,4424 lines, text taken from the
  GCC 3.3 distribution, sort.c compiled with GCC 2.95.4 and running on
  Debian 3.0r1 GNU/Linux, 2.4GHz Pentium 4, single pass with no
  temporary files and plenty of RAM.

  Since comparisons seem to be the bottleneck, perhaps the best
  algorithm to try next should be merge insertion.  See Knuth section
  5.3.1, who credits Lester Ford, Jr. and Selmer Johnson, American
  Mathematical Monthly 66 (1959), 387-389.

cp --recursive: perform dir traversals in source and dest hierarchy rather
  than forming full file names.  The latter (current) approach fails
  unnecessarily when the names become very long.

tail --p is now ambiguous

Remove suspicious uses of alloca (ones that may allocate more than
   about 4k)

Adapt these contribution guidelines for coreutils:
  http://sources.redhat.com/automake/contribute.html


Changes expected to go in, post-5.2.1:
======================================

  wc: add an option, --files0-from [as for du] to make it read NUL-delimited
    file name arguments from a file.

  dd patch from Olivier Delhomme

  Apply Andreas Gruenbacher's ACL and xattr changes

  Apply Bruno Haible's hostname changes

  test/mv/*: clean up $other_partition_tmpdir in all cases

  ls: when both -l and --dereference-command-line-symlink-to-dir are
  specified, consider whether to let the latter select whether to
  dereference command line symlinks to directories.  Since -l has
  an implicit --NO-dereference-command-line-symlink-to-dir meaning.
  Pointed out by Karl Berry.

  A more efficient version of factor, and possibly one that
  accepts inputs of size 2^64 and larger.

  dd: consider adding an option to suppress `bytes/block read/written'
  output to stderr.  Suggested here:
    http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=165045

  Pending copyright papers:
  ------------------------
  ls --color: Ed Avis' patch to suppress escape sequences for
    non-highlighted files

  getpwnam from Bruce Korb

  pb (progress bar) from Miika Pekkarinen

  ------------------------------
  Look into improving the performance of md5sum.
  `openssl md5' is consistently about 30% faster than md5sum on an idle
  AMD 2000-XP system with plenty of RAM and a 261 MB input file.
  openssl's md5 implementation is in assembly, generated by a perl script.

  On an AMD-64 system, using a 700MB file on a tmpfs file system
  (and enough RAM so that no actual disk reads were performed),
  GNU md5sum is slightly faster than `openssl md5', e.g.:

    2.38s user 0.38s system 100% cpu 2.756 total  (gnu md5sum)
    vs.
    2.52s user 0.34s system 100% cpu 2.869 total

  However, `openssl sha1' is about 5% faster than GNU sha1sum:

    3.32s user 0.33s system 99% cpu 3.653 total   (openssl sha1)
    3.45s user 0.39s system 99% cpu 3.843 total   (gnu sha1sum)

  The above are using the debian-sid (amd_64 alioth) binaries from
  coreutils-5.2.1.  When I compile the latest (coreutils-cvs) with
  gcc-4.0 -O3, I get slightly (2-3%) better sha1sum performance,
  and a ~7% *decrease* in performance for md5sum.  I suspect that
  with the right compiler options you can do much better.
  ------------------------------

Have euidaccess.m4 check for eaccess as well as euidaccess
If found, then do `#define euidaccess eaccess'.

Remove long-deprecated options like tail's --allow-missing

Add a distcheck-time test to ensure that every distributed
file is either read-only(indicating generated) or is
version-controlled and up to date.

Implement Ulrich Drepper's suggestion to use getgrouplist rather
  than getugroups.  This affects only `id', but makes a big difference
  on systems with many users and/or groups, and makes id usable once
  again on systems where access restrictions make getugroups fail.
  But first we'll need a run-test (either in an autoconf macro or at
  run time) to avoid the segfault bug in libc-2.3.2's getgrouplist.
  In that case, we'd revert to using a new (to-be-written) getgrouplist
  module that does most of what `id' already does.

remove `%s' notation:
  grep -E "\`%.{,4}s'" src/*.c

remove.c should never exit, yet may do so (see uses of EXIT_FAILURE)

remove or adjust chown's --changes option, since it
  can't always do what it currently says it does.

Adapt tools like wc, tr, fmt, etc. (most of the textutils) to be
  multibyte aware.  The problem is that I want to avoid duplicating
  significant blocks of logic, yet I also want to incur only minimal
  (preferably `no') cost when operating in single-byte mode.

Remove all uses of the `register' keyword

rm: add support for a -I option, like that from FreeBSD's rm:
   -I    Request confirmation once if more than three files are being
         removed or if a directory is being recursively removed.  This
         is a far less intrusive option than -i yet provides almost
         the same level of protection against mistakes.