]> diplodocus.org Git - nmh/log
nmh
8 years agosbr/m_maildir.c: Move prototypes from h/prototypes.h to new header.
Ralph Corderoy [Mon, 15 May 2017 23:30:59 +0000 (00:30 +0100)]
sbr/m_maildir.c: Move prototypes from h/prototypes.h to new header.

8 years agosbr/m_rand.c: Move prototype from h/prototypes.h to new header.
Ralph Corderoy [Mon, 15 May 2017 23:02:30 +0000 (00:02 +0100)]
sbr/m_rand.c: Move prototype from h/prototypes.h to new header.

8 years agoh/prototypes.h: Give sbr/lock_file.c's prototypes their own header.
Ralph Corderoy [Mon, 15 May 2017 22:54:05 +0000 (23:54 +0100)]
h/prototypes.h: Give sbr/lock_file.c's prototypes their own header.

8 years agosbr/utils.c: Delete unused upcase() function.
Ralph Corderoy [Mon, 15 May 2017 17:58:19 +0000 (18:58 +0100)]
sbr/utils.c: Delete unused upcase() function.

8 years agoh/utils.h: Move in PLURALS() macro from h/mh.h.
Ralph Corderoy [Mon, 15 May 2017 17:52:32 +0000 (18:52 +0100)]
h/utils.h: Move in PLURALS() macro from h/mh.h.

8 years agoh/prototypes.h: Remove duplicate prototype for pwd().
Ralph Corderoy [Mon, 15 May 2017 17:47:46 +0000 (18:47 +0100)]
h/prototypes.h: Remove duplicate prototype for pwd().

It is already in h/utils.h.

8 years agoh/utils.h: Move in nmh_init() prototype from h/prototypes.h.
Ralph Corderoy [Mon, 15 May 2017 17:39:45 +0000 (18:39 +0100)]
h/utils.h: Move in nmh_init() prototype from h/prototypes.h.

8 years agouip/mhshowsbr.h: Move in declarations from h/mhparse.h.
Ralph Corderoy [Mon, 15 May 2017 17:06:19 +0000 (18:06 +0100)]
uip/mhshowsbr.h: Move in declarations from h/mhparse.h.

Delete non-existent markerform parameter from show_all_messages()'s
comment.  Alter show_all_messages()'s definition's concatsw parameter to
match comment and prototype.

8 years agouip/mhshowsbr.c: Create header file with exports' declarations.
Ralph Corderoy [Mon, 15 May 2017 16:14:01 +0000 (17:14 +0100)]
uip/mhshowsbr.c: Create header file with exports' declarations.

Create a single set of externs for mhshowsbr.c's global variables.  Have
the other users of those use the new include file instead of their own
declarations.

8 years agouip/mhcachesbr.c: Move exported symbols to existing h/mhcachesbr.h.
Ralph Corderoy [Mon, 15 May 2017 15:09:28 +0000 (16:09 +0100)]
uip/mhcachesbr.c: Move exported symbols to existing h/mhcachesbr.h.

Remove the user's copies of the declarations.
A static struct swit caches[] was declared everywhere that included
h/mhcachesbr.h.  The larger number of includers causes `variable not
used' for `caches'.  Move that declaration to uip/mhcachesbr.c and have
an exported `cache_policy' symbol that points to it for the existing
users.

8 years agouip/mhoutsbr.c: Move exported prototype to new header file.
Ralph Corderoy [Mon, 15 May 2017 14:14:44 +0000 (15:14 +0100)]
uip/mhoutsbr.c: Move exported prototype to new header file.

Alter the callers to include the new header file instead of having their
own copy of the prototype.

8 years agosbr/m_mktemp.h: Move remaining prototypes from h/prototypes.h.
Ralph Corderoy [Mon, 15 May 2017 14:07:10 +0000 (15:07 +0100)]
sbr/m_mktemp.h: Move remaining prototypes from h/prototypes.h.

Gives the normal one header file showing interface exported by one C
file.

8 years agosbr/m_mktemp.h: Create with missing prototypes.
Ralph Corderoy [Mon, 15 May 2017 13:40:52 +0000 (14:40 +0100)]
sbr/m_mktemp.h: Create with missing prototypes.

Some of sbr/m_mktemp.c's functions didn't have prototypes in a header
file so callers were declaring prototypes themselves.  Have them include
the new header file instead.

8 years agosbr/utils.c: Use h/signals.h for setup_signal_handlers prototype.
Ralph Corderoy [Mon, 15 May 2017 13:31:36 +0000 (14:31 +0100)]
sbr/utils.c: Use h/signals.h for setup_signal_handlers prototype.

Deleted its own prototype for setup_signal_handlers() and added it to
the existing h/signals.h as it's defined in h/signals.c.

8 years agouip/picksbr.c: Use function prototypes for `nexus' functions.
Ralph Corderoy [Mon, 15 May 2017 13:23:53 +0000 (14:23 +0100)]
uip/picksbr.c: Use function prototypes for `nexus' functions.

Prototypes for ORaction(), etc., now state the parameters they expect.
Removed the args() macro, instead explicitly listing the arguments being
passed to functions.

8 years agoSpecify function parameters in prototypes, mainly void.
Ralph Corderoy [Mon, 15 May 2017 13:12:21 +0000 (14:12 +0100)]
Specify function parameters in prototypes, mainly void.

These prototypes are all in *.[cl] source compared with the *.h of the
earlier commit.  All but one had void added, that one's parameter was a
char pointer.

8 years agoh/prototypes.h: Remove scan_reset_m_getfld_state() prototype.
Ralph Corderoy [Mon, 15 May 2017 12:45:39 +0000 (13:45 +0100)]
h/prototypes.h: Remove scan_reset_m_getfld_state() prototype.

Function removed back in 2014 by 58eaf2c2.

8 years agoUse `void' in prototypes for functions with no parameters.
Ralph Corderoy [Mon, 15 May 2017 12:39:08 +0000 (13:39 +0100)]
Use `void' in prototypes for functions with no parameters.

The old-style `int foo()' still lived on in a few places.

8 years agoh/mhparse.h: Move mhfree.c's free_content prototype to new header.
Ralph Corderoy [Mon, 15 May 2017 12:29:28 +0000 (13:29 +0100)]
h/mhparse.h: Move mhfree.c's free_content prototype to new header.

It was the only thing of mhfree.c's in h/mhparse.h.

8 years agoMatch `NORETURN' function prototypes by adding it to definition.
Ralph Corderoy [Mon, 15 May 2017 12:25:23 +0000 (13:25 +0100)]
Match `NORETURN' function prototypes by adding it to definition.

8 years agoCreate and use header file for uip/mhfree.c's exports.
Ralph Corderoy [Mon, 15 May 2017 12:13:45 +0000 (13:13 +0100)]
Create and use header file for uip/mhfree.c's exports.

Remove all the duplicate inconsistent prototypes from mhfree.c's users.

8 years agotest/mhical/test-mhical: Force UTC timezone for mhical.
Ralph Corderoy [Mon, 15 May 2017 12:10:30 +0000 (13:10 +0100)]
test/mhical/test-mhical: Force UTC timezone for mhical.

Commit c3238c0e added a test but mhical's output depends on its
timezone;  force it to UTC for consistent results.

8 years agouip/mhfree.c: Make free_header() static.
Ralph Corderoy [Mon, 15 May 2017 11:52:23 +0000 (12:52 +0100)]
uip/mhfree.c: Make free_header() static.

8 years agouip/flist.c: Make locally defined and used functions static.
Ralph Corderoy [Mon, 15 May 2017 11:38:26 +0000 (12:38 +0100)]
uip/flist.c: Make locally defined and used functions static.

8 years agoSupport iCalendar event request files that don't end with a newline.
David Levine [Sun, 14 May 2017 21:27:23 +0000 (17:27 -0400)]
Support iCalendar event request files that don't end with a newline.

Such files do not conform to RFC 5545 ยง 3.1, which requires that each
contentline end with a CRLF.  But be liberal in what we accept.

8 years agoAdded explicit dependency to build sbr/icalparse.h before running LEX.
David Levine [Sun, 14 May 2017 17:08:03 +0000 (13:08 -0400)]
Added explicit dependency to build sbr/icalparse.h before running LEX.

"Fix" to commit 665dfc96.  The build had been relying upon the listed
order of files.

8 years agoHave post(1) report name of fileproc if it fails.
David Levine [Sun, 14 May 2017 16:18:35 +0000 (12:18 -0400)]
Have post(1) report name of fileproc if it fails.

8 years agoReplace calls to unputenv() with unsetenv(3).
Ralph Corderoy [Sun, 14 May 2017 22:17:36 +0000 (23:17 +0100)]
Replace calls to unputenv() with unsetenv(3).

8 years agoReplace calls to m_putenv() with setenv(3).
Ralph Corderoy [Sun, 14 May 2017 21:56:52 +0000 (22:56 +0100)]
Replace calls to m_putenv() with setenv(3).

8 years agosbr/dtimep.l: Remove redundant wrapping parenthesis in definitions.
Ralph Corderoy [Sun, 14 May 2017 18:35:13 +0000 (19:35 +0100)]
sbr/dtimep.l: Remove redundant wrapping parenthesis in definitions.

It's been decades since a lex(1) didn't follow POSIX and treat `{foo}'
as parenthesis surrounding foo's definition.  Other nmh lex files work
without the extra wrapping.

8 years agouip/mhstoresbr.c: Fix single-character nmh-storage bug.
Ralph Corderoy [Sun, 14 May 2017 11:53:33 +0000 (12:53 +0100)]
uip/mhstoresbr.c: Fix single-character nmh-storage bug.

If the profile component nmh-storage was a single character, e.g. the
relative directory `d', then it would be dropped from the output path so
instead of `d/foo' being written, `/foo' would probably fail.  I think
this was due to a faulty check for nmh-storage being `/'.  Add a test
for a non-/ single-character nmh-storage.

8 years agosbr/fmt_rfc2047.c: Simplify test for /^=\?./.
Ralph Corderoy [Sun, 14 May 2017 11:08:06 +0000 (12:08 +0100)]
sbr/fmt_rfc2047.c: Simplify test for /^=\?./.

8 years agoUse new PLURALS(n) macro instead of variety of tests.
Ralph Corderoy [Sun, 14 May 2017 10:23:33 +0000 (11:23 +0100)]
Use new PLURALS(n) macro instead of variety of tests.

Whether to output a plural noun in a message was decided by ternary
operators that tested n==1, n!=1, n>1, etc.  Make them all consistent by
using PLURALS(n).

8 years agouip/folder.c: Use `%s' for plural, not `%c'.
Ralph Corderoy [Sat, 13 May 2017 22:40:50 +0000 (23:40 +0100)]
uip/folder.c: Use `%s' for plural, not `%c'.

Instead of a ternary operator with 's' and ' ' as the outcomes, switch
the format specifier from `%c' to `%s' and use strings "s" and " ".

8 years agouip/folder.c: Rewrite plural test to common form.
Ralph Corderoy [Sat, 13 May 2017 17:39:06 +0000 (18:39 +0100)]
uip/folder.c: Rewrite plural test to common form.

The common form is a choice between "" and "s".  This one occurrence was
using " " and "s" so either output was a fixed width.  Alter the
printf(3) format string from `%s' to `%1s' to provide that instead.
Allows the new code to be altered by an upcoming `plural' change.

8 years agoMakefile.am: Alter long lists to be sorted, one entry per line.
Ralph Corderoy [Sat, 13 May 2017 17:35:15 +0000 (18:35 +0100)]
Makefile.am: Alter long lists to be sorted, one entry per line.

It was hard to see what files were included in some variables when
searching for the best place for new content.  Some variables had
definitions that were almost sorted, by weren't.  Switch to a
one-entry-per-line list, in `LC_ALL=C sort' order, documented at the
start of the file.  Makes it easy to run through the lists and see the
patterns.

8 years agoUse FENDNULL() instead of duplicate, locally defined, empty().
Ralph Corderoy [Sat, 13 May 2017 12:24:03 +0000 (13:24 +0100)]
Use FENDNULL() instead of duplicate, locally defined, empty().

Three files had an empty() that was defined using FENDNULL();  use the
latter directly.

8 years agoAdd FENDNULL(s): fends off NULL by giving an empty string instead.
Ralph Corderoy [Sat, 13 May 2017 12:13:18 +0000 (13:13 +0100)]
Add FENDNULL(s): fends off NULL by giving an empty string instead.

Use it in place of all the `foo->bar_xyzzy ? foo->bar_xyzzy : ""'.
Avoids the reader having to check first two of the ternary operands are
the same.

8 years agosbr/mf.c: Rename path to routepath to avoid h/mh.h clash.
Ralph Corderoy [Sat, 13 May 2017 11:56:58 +0000 (12:56 +0100)]
sbr/mf.c: Rename path to routepath to avoid h/mh.h clash.

Adding #include of h/mh.h brings two path global symbols together;
path() and char *path.  Rename the latter.  h/mh.h is needed by a future
commit.

8 years agoReplace `a == b ? 1 : 0' and similar with `a == b'.
Ralph Corderoy [Fri, 12 May 2017 22:29:31 +0000 (23:29 +0100)]
Replace `a == b ? 1 : 0' and similar with `a == b'.

8 years agoReplace `e ? 0 : 1' with `!e'.
Ralph Corderoy [Fri, 12 May 2017 22:14:43 +0000 (23:14 +0100)]
Replace `e ? 0 : 1' with `!e'.

Needs less thought to read.

8 years agosbr/icalparse.y: Replace max(BUFSIZ, 8192) with NMH_BUFSIZ.
Ralph Corderoy [Fri, 12 May 2017 21:45:34 +0000 (22:45 +0100)]
sbr/icalparse.y: Replace max(BUFSIZ, 8192) with NMH_BUFSIZ.

8 years agoUse existing macros min() and max() more.
Ralph Corderoy [Fri, 12 May 2017 21:42:35 +0000 (22:42 +0100)]
Use existing macros min() and max() more.

8 years agoClarified the phrasing added by commit 44ce2010.
David Levine [Fri, 12 May 2017 02:41:02 +0000 (22:41 -0400)]
Clarified the phrasing added by commit 44ce2010.

8 years agoAdded warning about when post(1) can't refile(1) a draft.
David Levine [Fri, 12 May 2017 00:20:16 +0000 (20:20 -0400)]
Added warning about when post(1) can't refile(1) a draft.

8 years agosbr/fmt_scan.c: Fix `foo%-42{bar}' right-justification bug.
Ralph Corderoy [Wed, 10 May 2017 11:13:28 +0000 (12:13 +0100)]
sbr/fmt_scan.c: Fix `foo%-42{bar}' right-justification bug.

The space padding was added at the start of the output buffer, not the
start of the component being formatted.  Caused by 92128dac's move to
dynamic allocation for fmt_scan()'s output.  Only shows if the component
isn't at the start of the buffer.  Expand existing
test/format/test-rightjustify to cover this.

8 years agoconfig/version.sh: Rewrite. Use uname(1), git-describe(1), and UTC.
Ralph Corderoy [Tue, 9 May 2017 22:10:06 +0000 (23:10 +0100)]
config/version.sh: Rewrite.  Use uname(1), git-describe(1), and UTC.

Not sure why it searched through PATH manually for uname(1) and
hostname(1).  uname and its -n option are POSIX so just use those.  It
was the preference over hostname anyway.

Use git-describe(1), not just git-branch(1), as it gives more detail,
including --dirty to show the built source differs from the commit.

Specify the format for the build date, and its timezone; +0000.

8 years agosbr/mf.c: Simplify isat(); it's /^ at /i.
Ralph Corderoy [Tue, 9 May 2017 21:48:11 +0000 (22:48 +0100)]
sbr/mf.c: Simplify isat();  it's /^ at /i.

This one's for all those RFC 733 emails out there.
You know who you are.

8 years agouip/inc.c: Use bool, not int, for some of the options.
Ralph Corderoy [Sun, 7 May 2017 22:30:46 +0000 (23:30 +0100)]
uip/inc.c: Use bool, not int, for some of the options.

Makes clear there's no special meaning to repeating these options more
than once.

8 years agosbr/netsec.c: Remove extraneous semicolon.
Ralph Corderoy [Sun, 7 May 2017 21:43:19 +0000 (22:43 +0100)]
sbr/netsec.c: Remove extraneous semicolon.

8 years agoSplit some comma-operator terms into separate statements.
Ralph Corderoy [Sun, 7 May 2017 21:35:41 +0000 (22:35 +0100)]
Split some comma-operator terms into separate statements.

Some uses of the comma operator weren't the idiomatic `p++, len--'.
Split its terms into two statements instead, e.g. `close(fd); fd = -1'.

8 years agouip/picksbr.c: Increase line-buffer size for "grep" action.
Ralph Corderoy [Sun, 7 May 2017 17:02:03 +0000 (18:02 +0100)]
uip/picksbr.c: Increase line-buffer size for "grep" action.

Alter LBSIZE from 1024 to NMH_BUFSIZ, e.g. 8192.  This dominates the
size of the char array used to hold a single unfolded header line when
evaluating a "grep" action, e.g. `-to foo' becomes `^to[ ^I]*:.*foo'.

8 years agouip/scansbr.c: Replace uses of SBUFSIZ with NMH_BUFSIZ.
Ralph Corderoy [Sun, 7 May 2017 13:01:48 +0000 (14:01 +0100)]
uip/scansbr.c: Replace uses of SBUFSIZ with NMH_BUFSIZ.

Now that the former is defined as the latter, the uses of SBUFSIZ were
all in the same expression and it simplifies from a ternary expression
to NMH_BUFSIZ.

8 years agouip/rcvdist.c: Replace SBUFSIZ with NMH_BUFSIZ.
Ralph Corderoy [Sun, 7 May 2017 12:57:39 +0000 (13:57 +0100)]
uip/rcvdist.c: Replace SBUFSIZ with NMH_BUFSIZ.

The former was defined as the latter and had only one proper use: to
size a char array.  The other uses should have been the sizeof operator
on that array, and now are.

8 years agouip/replsbr.c: Replace single use of SBUFSIZ with NMH_BUFSIZ.
Ralph Corderoy [Sun, 7 May 2017 12:56:40 +0000 (13:56 +0100)]
uip/replsbr.c: Replace single use of SBUFSIZ with NMH_BUFSIZ.

The former was defined as the latter;  not a lot of point having it.

8 years agoMake many m_getfld() buffer parameters NMH_BUFSIZ big.
Ralph Corderoy [Sun, 7 May 2017 12:03:01 +0000 (13:03 +0100)]
Make many m_getfld() buffer parameters NMH_BUFSIZ big.

char arrays were often stdio.h's BUFSIZ large, and that's 8192 with here
with glibc 2.25-1 on Linux x86_64, so using NMH_BUFSIZ is no different
as that's max(BUFSIZ, 8192).  But some were 256 or 512 with a local
SBUFSIZ macro and they caused scan(1) to truncate a field, or pick(1) to
not spot text because the field was truncated.

Delete bad-input/test-header's test for m_getfld()'s detection for a
header without a colon that's under the header-length NAMESZ limit, but
longer than the passed in buffer.  This no longer happens in scan as the
buffer is larger than the longest allowed header.

8 years agoconfigure.ac: Enable assert(3) by default.
Ralph Corderoy [Fri, 5 May 2017 12:11:15 +0000 (13:11 +0100)]
configure.ac: Enable assert(3) by default.

Revert 167e542b that disabled assert(3) by default after private email
discussion with David Levine.  I suggested it being enabled by default
in git so those running nmh from git were testing them for us, with it
being disabled as part of the release process so packagers continue with
it disabled.  David pointed out that would make the release
intentionally different from what had been long tested so the default is
now enabled assertions everywhere unless disabled by a packager.

8 years agomhstoresbr.c: Flip logic to simplify. Remove else after continue.
Ralph Corderoy [Mon, 1 May 2017 22:02:06 +0000 (23:02 +0100)]
mhstoresbr.c: Flip logic to simplify.  Remove else after continue.

Former in store_partial(), latter in parse_format_string().

8 years agomhlsbr.c: Flip parse()'s logic. Remove oneline()'s else after break.
Ralph Corderoy [Mon, 1 May 2017 21:57:44 +0000 (22:57 +0100)]
mhlsbr.c: Flip parse()'s logic.  Remove oneline()'s else after break.

8 years agodropsbr.c: Flip mbx_open()'s logic to simplify.
Ralph Corderoy [Mon, 1 May 2017 21:56:30 +0000 (22:56 +0100)]
dropsbr.c: Flip mbx_open()'s logic to simplify.

8 years agobase64.c: Remove some else after break. Flip logic to simplify.
Ralph Corderoy [Mon, 1 May 2017 21:47:38 +0000 (22:47 +0100)]
base64.c: Remove some else after break.  Flip logic to simplify.

8 years agomhfixmsg.c: Flip logic to simplify. Remove else after break.
Ralph Corderoy [Mon, 1 May 2017 21:46:38 +0000 (22:46 +0100)]
mhfixmsg.c: Flip logic to simplify.  Remove else after break.

8 years agopath.c: Simplify compath()'s flow; return, don't break twice.
Ralph Corderoy [Mon, 1 May 2017 21:36:30 +0000 (22:36 +0100)]
path.c: Simplify compath()'s flow;  return, don't break twice.

8 years agomhshowsbr.c: Remove else after break in iconv_start().
Ralph Corderoy [Mon, 1 May 2017 20:49:38 +0000 (21:49 +0100)]
mhshowsbr.c: Remove else after break in iconv_start().

8 years agomhparse.c: Remove a couple of else after continue.
Ralph Corderoy [Mon, 1 May 2017 20:46:20 +0000 (21:46 +0100)]
mhparse.c: Remove a couple of else after continue.

In openQuoted() and parse_header_attrs().

8 years agom_getfld.c: Remove else after break in m_unknown() and m_Eom().
Ralph Corderoy [Mon, 1 May 2017 20:43:36 +0000 (21:43 +0100)]
m_getfld.c: Remove else after break in m_unknown() and m_Eom().

8 years agopicksbr.c: Remove some else after break in plist().
Ralph Corderoy [Mon, 1 May 2017 20:42:59 +0000 (21:42 +0100)]
picksbr.c: Remove some else after break in plist().

8 years agofmt_rfc2047.c: Remove else after decode_rfc2047()'s break.
Ralph Corderoy [Mon, 1 May 2017 17:42:07 +0000 (18:42 +0100)]
fmt_rfc2047.c: Remove else after decode_rfc2047()'s break.

8 years agosortm.c: Flip get_fields()'s logic to simplify.
Ralph Corderoy [Mon, 1 May 2017 17:33:15 +0000 (18:33 +0100)]
sortm.c: Flip get_fields()'s logic to simplify.

8 years agoREADME.developers: Add pointers to Debian's Lintian's complaints.
Ralph Corderoy [Sun, 30 Apr 2017 14:25:20 +0000 (15:25 +0100)]
README.developers: Add pointers to Debian's Lintian's complaints.

8 years agoREADME.developers: Add blank lines for consistent headers.
Ralph Corderoy [Sun, 30 Apr 2017 14:14:46 +0000 (15:14 +0100)]
README.developers: Add blank lines for consistent headers.

8 years agoFix spelling in C comments.
Ralph Corderoy [Fri, 28 Apr 2017 23:01:58 +0000 (00:01 +0100)]
Fix spelling in C comments.

8 years agobvector_copy: Use bvector's tiny storage if big enough.
Ralph Corderoy [Fri, 28 Apr 2017 21:42:46 +0000 (22:42 +0100)]
bvector_copy: Use bvector's tiny storage if big enough.

Otherwise a source bvector that's using tiny storage has its bits
duplicated into a malloc'd area when the destination's tiny storage was
sufficient.

8 years agosbr/oauth.c: Remove const to avoid curl's debug_callback warning.
Ralph Corderoy [Fri, 28 Apr 2017 21:08:26 +0000 (22:08 +0100)]
sbr/oauth.c: Remove const to avoid curl's debug_callback warning.

The curl_debug_callback typedef doesn't use const for any of the
function's parameters.  On one of the compilation platforms here,
oauth.c's debug_callback having const specifiers for some parameters
causes compilation warnings, and -Werror stops the compilation.  Remove
them.

8 years agosbr/folder_read.c: Don't malloc() once per message.
Ralph Corderoy [Fri, 28 Apr 2017 16:12:31 +0000 (17:12 +0100)]
sbr/folder_read.c: Don't malloc() once per message.

Instead of struct msgs having a pointer to a malloc'd array of pointers,
each to a malloc'd struct bvector, 1+N, it now has a pointer to a
malloc'd array of struct bvector;  one malloc for all of them.  This
avoids the large number of calls to malloc() and free() that's linear
with the size of the folder.

But there are some downsides.  In order to step through an array of
struct bvector, code outside of sbr/vector.c needs to know the struct's
size.  The simplest way to do this is to make the struct's definition
public, with a comment that access should be through vector.c.

New functions are needed to initialise the content of an already
allocated bvector, and to finish with its content prior to deallocation.
bvector_create() and bvector_free() now also use these new functions.

Before, it was the array of pointers to bvector that would be realloc'd.
That doesn't work for the array of bvectors as they may contain pointers
to within themselves.  The solution is to malloc a new array and
bvector_copy() the ones to keep across, as folder_realloc() now does.
The other half of its logic that coped with growth at the end of the
array, has been deleted.  Also deleted, is the code to clear the
bvectors before and after the old ones as they start in that state.

8 years agovalgrind: Add suppression for dbm_open(3)'s write(2).
Ralph Corderoy [Fri, 28 Apr 2017 16:00:54 +0000 (17:00 +0100)]
valgrind: Add suppression for dbm_open(3)'s write(2).

It writes bytes from an area it mallocs, but doesn't set all of those
bytes first.  See with gdbm 1.13-1 on Arch Linux.

8 years agosbr/vector.c: Only allocate bvector storage for set bits.
Ralph Corderoy [Fri, 28 Apr 2017 11:45:26 +0000 (12:45 +0100)]
sbr/vector.c: Only allocate bvector storage for set bits.

Now that a pointer to all the bit storage doesn't escape vector.c,
there's no need to allocate storage, that's initialised to 0, just to
then clear a bit in it.  Only extend the bits's storage for set bits.
Remove the unused initial size parameter from bvector_create() as it
simplifies the implementation.

8 years agosbr/vector.c: Add bvector comments, tidy identifiers.
Ralph Corderoy [Fri, 28 Apr 2017 11:27:44 +0000 (12:27 +0100)]
sbr/vector.c: Add bvector comments, tidy identifiers.

8 years agosbr/vector.c: Rewrite BVEC_BYTES(n) macro to remove branch.
Ralph Corderoy [Fri, 28 Apr 2017 11:16:01 +0000 (12:16 +0100)]
sbr/vector.c: Rewrite BVEC_BYTES(n) macro to remove branch.

Use the idiom of integer truncation.

8 years agosbr/vector.c: Use new BVEC_BITS_BITS macro.
Ralph Corderoy [Fri, 28 Apr 2017 11:12:15 +0000 (12:12 +0100)]
sbr/vector.c: Use new BVEC_BITS_BITS macro.

Factors out a common expression from a few other places.

8 years agosbr/vector.c: Zero the growth with memset(3), not loop.
Ralph Corderoy [Wed, 26 Apr 2017 23:14:28 +0000 (00:14 +0100)]
sbr/vector.c: Zero the growth with memset(3), not loop.

When the resize functions grow the vectors, as they always do, zero the
new slots with memset(3) rather than a for loop.  Particularly of note
for the bit vector where it was bvector_clear()ing one bit at a time.
Although a NULL pointer needn't have a representation of all-zero bits,
the code was already assuming that, e.g. on the initial allocation, so
we're no worse off.

8 years agosbr/vector.c: Change bvector_bits() to return first word.
Ralph Corderoy [Wed, 26 Apr 2017 12:41:27 +0000 (13:41 +0100)]
sbr/vector.c: Change bvector_bits() to return first word.

Rename it to bvector_first_bits() to represent its new behaviour.  It
has only one caller that uses it to produce debug.  Tighten the API so
other callers don't get access to the location of the bvector's bits.

8 years agosbr/vector.c: Embed initial vector storage in header struct.
Ralph Corderoy [Tue, 25 Apr 2017 22:21:21 +0000 (23:21 +0100)]
sbr/vector.c: Embed initial vector storage in header struct.

Instead of a malloc(3)'d struct bvector having a pointer to a separately
malloc'd area for the bits in the normal case of the default initial
size, have non-malloc'd storage in the struct itself, and set the
pointer to that.  It's two unsigned longs, which is less than the
previous default of 256 bits for struct bvector, but still double the
pre-bvector norm of one word on 32 and 64-bit architectures.

This halves the mallocs needed to create a struct bvector in the common
case, but does mean that embedded memory is wasted should it not be
enough.  That's probably an unusual case.  It also means derefencing the
pointer to the bits probably hits the same cache line.

Have separate initial sizes for string and int vectors.

8 years agosbr/vector.c: Remove `vec' argument from BVEC_OFFSET(), etc.
Ralph Corderoy [Mon, 24 Apr 2017 23:20:08 +0000 (00:20 +0100)]
sbr/vector.c: Remove `vec' argument from BVEC_OFFSET(), etc.

A few macros took a struct bvector pointer as their first argument just
to get the sizeof one of its fields.  This can be done with a NULL
pointer so remove that argument from all of them.

8 years agosbr/vector.c: Delete unused bvector_maxsize() and ivector_size().
Ralph Corderoy [Mon, 24 Apr 2017 22:47:46 +0000 (23:47 +0100)]
sbr/vector.c: Delete unused bvector_maxsize() and ivector_size().

May as well keep the proffered interface as small as possible so the
implementation can make greater assumptions knowing some internal
details aren't available.

8 years agosbr/vector.c: Move assert(3)s into bvector_create().
Ralph Corderoy [Mon, 24 Apr 2017 21:58:56 +0000 (22:58 +0100)]
sbr/vector.c: Move assert(3)s into bvector_create().

Rather than asserting on every bit-related operation, assert just when
creating the vector.  Whether the assertion is true is decided at
compile time, so it only needs checking once;  once for every vector is
an easy compromise.

8 years agosbr/vector.c: Replace Nbby with <limits.h>'s CHAR_BIT.
Ralph Corderoy [Mon, 24 Apr 2017 21:08:09 +0000 (22:08 +0100)]
sbr/vector.c: Replace Nbby with <limits.h>'s CHAR_BIT.

POSIX mandates CHAR_BIT is 8;  see stdint.h(0p).

8 years agosbr/vector.c: calloc(3) rather than malloc(3) and memset(3).
Ralph Corderoy [Mon, 24 Apr 2017 20:19:56 +0000 (21:19 +0100)]
sbr/vector.c: calloc(3) rather than malloc(3) and memset(3).

calloc(3) tells libc upfront that the memory needs to be zeroed rather
than giving it the news later with memset.  Perhaps this allows it to
allocate from a CoW page of zero bytes, or it benefits from some other
way in memset not having to clear each bit, but it knocks about 15% off
the wall-clock time and the number of library calls under ltrace(1).

8 years agosbr/folder_read.c: Don't clear newly created bvectors.
Ralph Corderoy [Mon, 24 Apr 2017 20:13:52 +0000 (21:13 +0100)]
sbr/folder_read.c: Don't clear newly created bvectors.

The newly created bvectors are already clear;  don't loop clearing each
again with memset(3).

8 years agosbr/fmt_scan.c: Only wcwidth(3) a valid mbtowc(3) result.
Ralph Corderoy [Tue, 25 Apr 2017 23:14:26 +0000 (00:14 +0100)]
sbr/fmt_scan.c: Only wcwidth(3) a valid mbtowc(3) result.

The assert(3) added by 80a9e99f7078199500d2d53c8d77d1b92af06fbc is
failing, but not reproducibly.  It's probable that mbtowc() is returning
a negative, and not altering wide_char, leaving it as random stack
content.  Taking its wcwidth() then sometimes also returns negative,
causing the assert() failure.  Initialising wide_char before the call
isn't a solution as it isn't documented if it's modified to an invalid
value on an error return.

Instead, delay calculating the wcwidth() until after the possible
substitution of "?".  Leave the assert() in place.

8 years agoRemove return statements at end of void functions.
Ralph Corderoy [Sun, 23 Apr 2017 15:17:09 +0000 (16:17 +0100)]
Remove return statements at end of void functions.

8 years agotws.h: Remove ADJUST_NUMERIC_ONLY_TZ_OFFSETS_WRT_DST.
Ralph Corderoy [Sun, 23 Apr 2017 15:10:59 +0000 (16:10 +0100)]
tws.h: Remove ADJUST_NUMERIC_ONLY_TZ_OFFSETS_WRT_DST.

Code that's conditional on its definition has been kept;  just the tests
removed.  The comment explaining its purpose has been kept, just without
the "If defined...".

8 years agoUse C's `++', `+=', etc., not the longhand.
Ralph Corderoy [Sun, 23 Apr 2017 14:48:53 +0000 (15:48 +0100)]
Use C's `++', `+=', etc., not the longhand.

8 years agoRemove unneeded incomplete, member-less `struct tag;'.
Ralph Corderoy [Sun, 23 Apr 2017 14:19:33 +0000 (15:19 +0100)]
Remove unneeded incomplete, member-less `struct tag;'.

The forward declaration isn't needed to typedef based on the struct.

8 years agomh.h: Compact the reserved sequence bit-masks.
Ralph Corderoy [Sun, 23 Apr 2017 13:47:20 +0000 (14:47 +0100)]
mh.h: Compact the reserved sequence bit-masks.

Removes the gap created by axing DELETED.

8 years agomh.h: Remove unused MODIFIED and DELETED macros.
Ralph Corderoy [Sun, 23 Apr 2017 13:36:57 +0000 (14:36 +0100)]
mh.h: Remove unused MODIFIED and DELETED macros.

MODIFIED was a folder attribute used by msh(1), and DELETED a message
attribute that may have last been used in the 1980s.
Adjust corresponding FBITS and MBITS definitions.

8 years agomd5.h: Remove unused UINT2 typedef for unsigned short.
Ralph Corderoy [Sun, 23 Apr 2017 13:36:47 +0000 (14:36 +0100)]
md5.h: Remove unused UINT2 typedef for unsigned short.

8 years agoap.man, dp.man: Add fmttest(1) to SEE ALSO section.
Ralph Corderoy [Sun, 23 Apr 2017 12:50:02 +0000 (13:50 +0100)]
ap.man, dp.man: Add fmttest(1) to SEE ALSO section.

8 years agotws.h: Remove zero-valued TW_SNIL macro, used once.
Ralph Corderoy [Sun, 23 Apr 2017 12:33:57 +0000 (13:33 +0100)]
tws.h: Remove zero-valued TW_SNIL macro, used once.

Macro TW_SNIL was the zero value for a two-bit field.  It was only used
once, and that is better ordered to check for bits set with none set
being the last, else, case, rather than in the middle.  Kept the
behaviour of undefined value 3 being treated as TW_SIMP, 2.

8 years agotws.h: Remove redundant TW_SZONE and TW_SZNIL bit-masks.
Ralph Corderoy [Sun, 23 Apr 2017 12:18:20 +0000 (13:18 +0100)]
tws.h: Remove redundant TW_SZONE and TW_SZNIL bit-masks.

TW_SZONE is a mask of one bit that can either be 0, TW_SZNIL, or set,
TW_SZEXP.  Rather than three macros for the single  bit, have one,
TW_SZEXP.  Simplifies the test too when the mask is known to be one-bit
wide.