empserver
5 years agoshow: Extend show item to show the power value
Markus Armbruster [Fri, 27 May 2016 19:40:11 +0000 (21:40 +0200)]
show: Extend show item to show the power value

Also update "info power" to point to "show item" instead of the
formerly hardcoded values.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agopower: Make item power value configurable
Markus Armbruster [Fri, 27 May 2016 19:18:47 +0000 (21:18 +0200)]
power: Make item power value configurable

Custom games may want to tweak how items contribute to the power
factor, in particular when products are also customized.  Add ichrstr
member i_power and item selector power for that.

"info power" doesn't reflect this change, yet.  It'll be updated in
the next commit.

The current item power values are problematic.  This will be addressed
later.

For straightforward configurations, reasonable item power values could
perhaps be derived from the configuration automatically.  However,
this is surprisingly hard in the general case: since producing things
should not decrease power, the efficiency of processing products into
other products needs to be considered, and estimating these
efficiencies can be difficult.

Deities can create multiple products making the same item, or multiple
sector types with the same product, but different process efficiency
(sect-chr selector peffic).  Providing differently efficient ways to
make the same item can be reasonable when the sector types involved
have different terrain.  To average them, you'd need to know the map.

The stock game has one example: gold mines produce dust with 100%
process efficiency, mountains produce it with 75%.  Mountains are
normally rare enough not to matter.

Level p.e. (product selectors nlmin, nllag) may have to be considered.
In the stock game, level p.e. variations are minor, because it reaches
0.9 pretty quickly.  In games where it doesn't, you might have to
increase the power value of the product.

Resources (sect selectors min, gold, fert, ocontent, uran) and
resource depletion (product selectors nrndx and nrdep) further
complicate things: you might want to increase the power value of
products depending on unusually scarce resources, but you can't know
what's scarce without understanding the map.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agopower: Use ship, plane, land unit tech instead of nation's
Markus Armbruster [Fri, 27 May 2016 18:08:39 +0000 (20:08 +0200)]
power: Use ship, plane, land unit tech instead of nation's

Actual abilities of ships, planes and land units depend almost
completely on the individual unit's tech, not the nation's tech.  The
power factor should reflect that.

The power value of a unit is of the form

    base value * (20 + nation's tech level) / 500

Change it to

    base value * (20 + unit's tech level) / 500

Note that a plane's base value still depends on the nation's tech
level.  This commit merely makes the absurdity stand out a bit more.
To be fixed later.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agopower: Saner power value for tech, particularly at low tech
Markus Armbruster [Fri, 27 May 2016 17:58:58 +0000 (19:58 +0200)]
power: Saner power value for tech, particularly at low tech

In the old times, power didn't consider tech at all.  Chainsaw's
option NEWPOWER (mandatory since v4.2.14, on by default before)
changed this dramatically: the power factor gets multiplied by
max(1, tech) / 500.

In the early game, small absolute tech differences yield large power
factor differences.  For instance, if country A has tech level 10, and
B has 5, then A gets a factor two boost.

As the game progresses, tech differences between viable countries tend
to grow, but only slowly.  The influence on power diminishes.  For
instance, if C has tech level 270 and D has 240 (quite a respectable
tech lead), then C gets a modest 1.125x boost over D.

Change the factor to (20 + tech) / 500.  Now A's advantage is only
1.2, and C's is 1.115.

You might think that's rather low.  However, tech is not power unless
you project it, and then it manifests itself as sectors, population
and other stuff power counts.

The same tech term occurs in plane power, except with just tech
instead of max(1, tech) .  Change it there as well, for consistency.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agoinfo/power: Rewrite power formula for clarity
Markus Armbruster [Fri, 27 May 2016 15:36:35 +0000 (17:36 +0200)]
info/power: Rewrite power formula for clarity

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agopower: Drop the RES_POP factor
Markus Armbruster [Fri, 27 May 2016 14:58:04 +0000 (16:58 +0200)]
power: Drop the RES_POP factor

If option RES_POP is enabled, the power factor is multiplied by a
"research factor" of 1.0 + maxpop / 10000.0, where maxpop is the
maximum population of a mine sector.

Back when this code was written (Chainsaw 3), all sectors had the same
population limit, so using a mine sector was as good as any.  Since
then, it has become configurable, and the stock game has both sector
types with lower (mountains, plains) and with higher (cities)
population limits.

Space for people is worth considering for power, but multiplying total
power by a fudge factor based on the most common sector type's maximum
population is silly.  Drop it.

Adjusting each sector's value for maximum population would make more
sense, with and without RES_POP.  Perhaps later.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agopower: Simplify tallying the power value of items
Markus Armbruster [Fri, 27 May 2016 13:42:17 +0000 (15:42 +0200)]
power: Simplify tallying the power value of items

Item power value is basically number of items times the item power
weight.  For some item types, we add up the item numbers, then apply
the weight.  For lcms and hcms, we apply the weight, then add up the
weighted numbers.

Adopt the latter method for all types: change addtopow() to tally the
power value for all types instead of just lcms and hcms, and drop
gen_power()'s item power value computation.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agoconfig: Generalize unit build materials storage
Markus Armbruster [Fri, 27 May 2016 12:17:08 +0000 (14:17 +0200)]
config: Generalize unit build materials storage

Use a single array member instead of multiple scalar members.  Only
the array elements that replace scalar members are can be non-zero for
now.

This is a first step to permitting more build materials.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agoconfig: Make work to build units independently configurable
Markus Armbruster [Fri, 27 May 2016 11:39:11 +0000 (13:39 +0200)]
config: Make work to build units independently configurable

The work required for build and repairs is traditionally a function of
build materials: 20 + lcm + 2*hcm for ships, planes and land units,
and (lcm + 2*hcm + oil + rad)/5 for nukes.  Make it independently
configurable instead, via new ship-chr, plane-chr, land-chr, nuke-chr
selector bwork, backed by new struct mchrstr member m_bwork, struct
plchrstr member pl_bwork, struct lchrstr member l_bwork, struct
nchrstr member n_bwork.  Keep the required work exactly the same for
now.

Clients that compute work from materials need to be updated.  Easy,
since build work is now exposed in xdump.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agotests/update: New; exercises the update
Markus Armbruster [Tue, 28 Jun 2016 18:29:18 +0000 (20:29 +0200)]
tests/update: New; exercises the update

Notable gaps in its coverage are fallout, most of guerrilla, delivery,
distribution, ALL_BLEED and LOSE_CONTACT.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agoedit: Add editing of land unit plague
Markus Armbruster [Tue, 28 Jun 2016 17:34:27 +0000 (19:34 +0200)]
edit: Add editing of land unit plague

Add edit u keys 'A' for plague stage, and 'b' for plague time.
Admittedly unobvious, but at least they match edit s keys 'a' and 'b'.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agotests/version: New; exercises config introspection commands
Markus Armbruster [Sun, 29 May 2016 17:27:13 +0000 (19:27 +0200)]
tests/version: New; exercises config introspection commands

Exercise version, show and xdump, except for xdump of game state.

The xdump part is mostly factored out of tests/smoke.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agoship.config plane.config: Drop obsolete notes on auto-set flags
Markus Armbruster [Fri, 27 May 2016 08:58:58 +0000 (10:58 +0200)]
ship.config plane.config: Drop obsolete notes on auto-set flags

Flags monkey-patching is gone since commit c0c5822 and commit a4a25df,
v4.3.33.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agoinfo/launch: Clean up description of satellites a bit
Markus Armbruster [Thu, 19 May 2016 19:25:49 +0000 (21:25 +0200)]
info/launch: Clean up description of satellites a bit

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agoplnsub: Make takeoff/landing in mountains consistent
Markus Armbruster [Tue, 17 May 2016 20:03:40 +0000 (22:03 +0200)]
plnsub: Make takeoff/landing in mountains consistent

One-way sorties (fly, recon and sweep) reject mountain destinations
with a "Nowhere to land" message.  However, planes can land there just
fine when they return to base (bomb, drop, paradrop, missions).
Already inconsistent in BSD Empire 1.1.

Fix the inconsistency by changing pln_where_to_land() to permit only
helicopters to land in mountains, and pln_airbase_ok() to permit only
helicopters and missiles to take off there, i.e. reject fixed-wing
aircraft.

The flying commands now reject fixed-wing planes based in mountains
with an "is in a mountain and can't take off" message.

Commands flying to a mountain now select only helicopters and silently
ignore the rest, exactly like they select only VTOL planes for flying
to a non-airfield.  If no planes can be selected, the command fails
with a "No planes could be equipped" message.  This is admittedly less
clear than the "Nowhere to land" message we got before.

Missions now ignore fixed-wing planes based in mountains, exactly like
they ignore non-VTOL planes outside airfields.  This may make players
wonder why the fixed-wing VTOL planes they transported up that
mountain don't obey missions.  Missions are always quiet unless they
execute.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agoplnsub: Add missing newline to two messages
Markus Armbruster [Mon, 16 May 2016 16:43:07 +0000 (18:43 +0200)]
plnsub: Add missing newline to two messages

The two "while it is carrying a nuclear weapon" messages lack
newlines.  Add them.  Screwed up in commit a269cdd, v4.3.23.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agobomb: Disallow bombing spy units
Markus Armbruster [Sat, 14 May 2016 20:15:21 +0000 (22:15 +0200)]
bomb: Disallow bombing spy units

When bombing land units, the bombers get a chance to spot spies.  They
can target one even when it wasn't spotted.  This makes no sense.
Screwed up when spy units were added in 4.0.0.  Hide them completely.
They can still be killed via collateral damage.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agoRevert "subs: Add unitsatxy() parameter only_count"
Markus Armbruster [Sat, 14 May 2016 20:13:57 +0000 (22:13 +0200)]
Revert "subs: Add unitsatxy() parameter only_count"

This reverts commit 9b33a4c59824a249ee6f04a94045bcb1f93f4f77.

Parameter only_count was introduced so would_abandon() could use
unitsatxy(), but that was a flawed idea, fixed in the previous commit.
No callers passing non-zero remain, so get rid of it.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agomarch: Fix check for sector abandonment
Markus Armbruster [Sat, 14 May 2016 20:00:36 +0000 (22:00 +0200)]
march: Fix check for sector abandonment

sct_prewrite() makes an owned sector revert to the deity when there
are no civilians, military or own land units.

would_abandon() tries to predict that, but gets it wrong: it ignores
land units that evade spy detection or are loaded on ships, and it
fails to ignore land units loaded on land units marching out.

Broken in commit 7c1b166, v4.3.33.  Fix by counting manually rather
than with unitsatxy().

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agofire: Fix damage and ammunition use of return fire
Markus Armbruster [Sat, 14 May 2016 17:53:02 +0000 (19:53 +0200)]
fire: Fix damage and ammunition use of return fire

quiet_bigdef() runs for each attacker.  It lets each eligible defender
fire at most once.  The first time a defender is eligible, it fires
and is saved in the list of defenders, along with its firing damage.
If it's eligible again for a later attacker, it's found in the list of
defenders, and the damage is reused.  The list of defenders searched
with search_flist().  Unfortunately, search_flist() compares only uid,
not type, and therefore can return a previously found defender of
another type.

If there are multiple attackers and multiple defenders with the same
uid, total damage can be off, damage can be spread to attackers out of
range, and defenders may not be charged shells.  Abuse is possible,
but complicated to set up, and probably not worth the trouble.

Broken in commit f89edc7, v4.3.12.  Fix by comparing the type as well.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agollook: Drop useless "spy loaded" conditional
Markus Armbruster [Sat, 14 May 2016 14:51:39 +0000 (16:51 +0200)]
llook: Drop useless "spy loaded" conditional

The "loaded on ship" condition was useless from the start (v4.2.0).
The "loaded on land" condition became useless in commit 45d090b,
v4.3.28.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agoUpdate copyright notice
Markus Armbruster [Sat, 2 Jan 2016 08:19:50 +0000 (09:19 +0100)]
Update copyright notice

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
5 years agocommon/rdsched: Document why we need _XOPEN_SOURCE
Markus Armbruster [Fri, 25 Dec 2015 15:56:12 +0000 (16:56 +0100)]
common/rdsched: Document why we need _XOPEN_SOURCE

It's for strptime().

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoman/empire: Trim unwanted space in synopsis
Markus Armbruster [Sun, 13 Dec 2015 08:18:54 +0000 (09:18 +0100)]
man/empire: Trim unwanted space in synopsis

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agodoc/contributing: Fix git format-patch topic branch example
Markus Armbruster [Sun, 13 Dec 2015 07:27:26 +0000 (08:27 +0100)]
doc/contributing: Fix git format-patch topic branch example

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agotravis: Enable OS X
Markus Armbruster [Sun, 29 Nov 2015 13:53:40 +0000 (14:53 +0100)]
travis: Enable OS X

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agonavigate march: Fix abort not to wipe out concurrent updates
Markus Armbruster [Sun, 29 Nov 2015 10:30:45 +0000 (11:30 +0100)]
navigate march: Fix abort not to wipe out concurrent updates

When the player aborts the command at the movement prompt, we write
back stale ships or land units, triggering a generation oops.  Any
updates made by other threads meanwhile are wiped out, triggering a
seqno mismatch oops.

Broken in commit 24000b4, v4.3.33.  Fix by restoring the lost
shp_nav_stay_behind() and lnd_mar_stay_behind() calls.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agotests/navi-march: Cover abort at movement prompt
Markus Armbruster [Sun, 29 Nov 2015 09:53:14 +0000 (10:53 +0100)]
tests/navi-march: Cover abort at movement prompt

This exposes generation oopses.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agomarch: Fix concurrent updates at sector abandon prompt
Markus Armbruster [Sun, 29 Nov 2015 09:37:40 +0000 (10:37 +0100)]
march: Fix concurrent updates at sector abandon prompt

When the player declines to abandon a sector, we write back stale land
units, triggering a generation oops.  Any updates made by other
threads meanwhile are wiped out, triggering a seqno mismatch oops.

The culprit is lnd_abandon_askyn(): when the player declines, it
returns without calling check_sect_ok(), check_land_ok().  Broken in
commit 7c1b166, v4.3.33.  Fix it.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agorecvclient: Track potential yield on input
Markus Armbruster [Sun, 29 Nov 2015 09:20:52 +0000 (10:20 +0100)]
recvclient: Track potential yield on input

recvclient() calls ef_make_stale() only when it does actual I/O, via
io_output() and io_input().  Missed in commit 2fa5f652, v4.3.24.  Call
it directly when it doesn't do actual I/O.

This makes navi-march-test expose a bug in march: when the player
declines to abandon a sector, we write back stale land units,
triggering a generation oops.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoconfigure: Use -fstack-protector-strong when available
Markus Armbruster [Sun, 22 Nov 2015 09:38:24 +0000 (10:38 +0100)]
configure: Use -fstack-protector-strong when available

Testing whether the compiler supports it is a bit tricky.

The obvious AX_APPEND_COMPILE_FLAGS([-fstack-protector-strong])
doesn't suffice, since some ports of the GNU toolchain reportedly pass
this test, then fail to link.  That's because the compiler accepts the
flag, duly emits references to helper code in libc, but libc doesn't
provide, and linking fails.

Instead, use AX_APPEND_LINK_FLAGS with an input source that makes the
compiler emit the extra stack checking code.  This requires the latest
version from the autoconf-archive, so update m4/ax* to commit e3d948b.
Also update m4/my_append_compile_flags.m4 to keep it in sync with
upstream's ax_append_compile_flags.m4.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoscripts/savecore: Report nicely when there's no core dump
Markus Armbruster [Sun, 15 Nov 2015 07:40:36 +0000 (08:40 +0100)]
scripts/savecore: Report nicely when there's no core dump

When savecore can't find a core dump, it reports something like

    ls: cannot access core.*: No such file or directory

to stderr, and fails.  If privlog is set, it also mails out a "Could
not save core dump" note.

Suppress the error message, and mail out "Could not find core dump to
save" instead.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoconfigure: Use -fno-strict-aliasing -fno-strict-overflow
Markus Armbruster [Sun, 18 Oct 2015 15:48:15 +0000 (17:48 +0200)]
configure: Use -fno-strict-aliasing -fno-strict-overflow

Contemporary compilers can squeeze out some extra performance by
assuming the program never executes code that has undefined behavior
according to the C standard.  Unfortunately, this can break programs.
Pointing out that these programs are non-conforming is as correct as
it is unhelpful, at least as long as the compiler is unable to
diagnose the non-conformingness.

Since keeping our programs working is a lot more important to us than
running them as fast as possible, forbid some assumptions that are
known to break real-world programs:

* Aliasing: perfectly clean programs don't engage in type-punning, and
  perfectly conforming programs do it only in full accordance with the
  standard's (subtle!) aliasing rules.  Neither kind of perfection is
  realistic for us, therefore -fno-strict-aliasing.

* Signed integer overflow: perfectly clean programs won't ever do
  signed integer arithmetic that overflows.  This is an imperfect
  program, therefore -fno-strict-overflow.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agotests: Enable GNU libc memory allocation error checking
Markus Armbruster [Sun, 27 Sep 2015 07:56:21 +0000 (09:56 +0200)]
tests: Enable GNU libc memory allocation error checking

MALLOC_CHECK_=3 makes glibc check for memory allocation programming
errors.  It's the factory default, but set it anyway just in case
someone disabled it for speed.

Non-zero MALLOC_PERTURB_ makes glibc wipe memory value on allocation
and deallocation.  The actual value determines the bit pattern.  Set
it to the value of environment variable EMPIRE_CHECK_MALLOC_PERTURB or
else a pseudo-random number, and record it in sandbox/malloc-perturb.

See mallopt(3) for more information.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agonavigate march: Plug memory leaks
Markus Armbruster [Sun, 20 Sep 2015 13:24:22 +0000 (15:24 +0200)]
navigate march: Plug memory leaks

When the player aborts the command at the movement prompt, or declines
to abandon a sector, unit_move() returns without freeing the list.
Found with valgrind.  Broken in commit 24000b4 and commit 7c1b166,
both v4.3.33.

Free the list on these returns, too.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agonavigate march retreat lretreat: Fix read beyond buffer
Markus Armbruster [Sun, 20 Sep 2015 11:17:33 +0000 (13:17 +0200)]
navigate march retreat lretreat: Fix read beyond buffer

shp_nav_gauntlet() and lnd_mar_gauntlet() read beyond the list head
when the list is empty.  The values read aren't used then.  Could
conceivably crash the server anyway, but it's unlikely.

Empty list happens when shp_nav_dir(), lnd_mar_dir() empty the list
and return zero.  Broken in commit beedf8d, v4.3.33.  Occurs in
navi-march-test (since the last commit) and in retreat-test.

Change shp_nav_dir() and lnd_mar_dir() to return one then.  For
additional safety, make shp_nav_gauntlet() and lnd_mar_gauntlet() oops
on empty list and recover safely.

I think I originally found this bug with -fsanitize, but I've since
upgraded, and I can't diagnose it that way anymore.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agotests/navi-march: Cover running out of mobility completely
Markus Armbruster [Sun, 20 Sep 2015 10:30:33 +0000 (12:30 +0200)]
tests/navi-march: Cover running out of mobility completely

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agobomb drop fly paradrop recon sweep: Fix read before array
Markus Armbruster [Sun, 26 Jul 2015 18:28:34 +0000 (20:28 +0200)]
bomb drop fly paradrop recon sweep: Fix read before array

The code computing the length of the flight path checks whether the
path ends with 'h'.  When getpath() returns an empty path, it accesses
flightpath[-1].  This could set the length to -1 (unlikely), or crash
(even less likely).  The former could be abused to gain mobility for
sufficiently inefficient or short-ranged planes.  Found with valgrind.

Broken in commit 404a76f7, v4.3.27.

Historically, getpath() could return paths with or without 'h', and
the check was necessary.  It returned an empty path only when the
player gave no input, aborting the command.  When the player entered
the assembly point's coordinates, it returned "h".

Commit 404a76f7 accidentally changed it to return "" then.  Also broke
flying to the assembly point's coordinates.  Commit 0f1e14f (v4.3.31)
fixed that part by changing getpath()'s contract: always return paths
without 'h' ("" simply means empty path), and return NULL on invalid
input, including no input.

The flawed check is superfluous since then.  Drop it.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoAvoid shifting into sign bit
Markus Armbruster [Sun, 26 Jul 2015 16:00:27 +0000 (18:00 +0200)]
Avoid shifting into sign bit

It's undefined behavior.  Found with gcc -fsanitize=undefined.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoConvert run-time to build-time assertion
Markus Armbruster [Sat, 4 Jul 2015 14:19:16 +0000 (16:19 +0200)]
Convert run-time to build-time assertion

There's just one, in show_product().

Use new BUILD_ASSERT() there, because its contract is even simpler
than BUILD_ASSERT_ONE()'s.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoProvide proper build-time assertions for NSC_SITYPE()
Markus Armbruster [Sat, 4 Jul 2015 12:53:20 +0000 (14:53 +0200)]
Provide proper build-time assertions for NSC_SITYPE()

We want to cause a diagnostic when NSC_SITYPE()'s argument isn't
implemented.  Commit aa6ad9d's solution is to have the macro expand
into 1/0 then.  Works with GCC, but Clang always warns "division by
zero is undefined".

The better, portable way to conditionally break the build is an array
type with a size that's negative when the build should fail, else
positive.  Implement that wrapped in a sizeof() to make it an
expression as macro BUILD_ASSERT_ONE(), and use it in NSC_SITYPE().

No more warnings from Clang 3.5.0.  GCC still produces its "may be
used uninitialized" false positives.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agosubs: Don't squash telegrams together when time goes backwards
Markus Armbruster [Sat, 4 Jul 2015 12:31:28 +0000 (14:31 +0200)]
subs: Don't squash telegrams together when time goes backwards

We've always squashed them when the time difference is smaller than
TEL_SECONDS, regardless of sign.  This involves passing the difference
to abs(), implicitly casting from time_t to int, which triggers a
Clang warning.

I could clean this up to get rid of the warning, but time should never
go backwards, and trying to make things prettier when it does isn't
worthwhile.  Simply drop the abs().

While there, drop the function comment.  It's been inaccurate since
Empire 3 dropped mail.c, and bogus since commit 17223e8 (v4.3.29)
added tel_cont.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoinfo/Nuke-types: Document show columns avail, res, abilities
Markus Armbruster [Thu, 2 Jul 2015 07:40:34 +0000 (09:40 +0200)]
info/Nuke-types: Document show columns avail, res, abilities

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoinfo/Unit-types: Belatedly remove capability xlight
Markus Armbruster [Thu, 2 Jul 2015 07:19:48 +0000 (09:19 +0200)]
info/Unit-types: Belatedly remove capability xlight

L_XLIGHT was replaced in commit e28c14f, v4.3.0.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoinfo/Plane-types: Belatedly remove stealth and half-stealth
Markus Armbruster [Thu, 2 Jul 2015 07:17:29 +0000 (09:17 +0200)]
info/Plane-types: Belatedly remove stealth and half-stealth

P_X and P_H were removed in commit 61233e4, v4.3.23.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoinfo/Ship-types: Belatedly remove capability spy
Markus Armbruster [Thu, 2 Jul 2015 07:14:14 +0000 (09:14 +0200)]
info/Ship-types: Belatedly remove capability spy

M_SPY was removed in commit 498d9fb, v4.3.0.  It never did anything.

Reported-by: Harald Katzer
Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoinfo: Belatedly update for change of stop prefix to '!'
Markus Armbruster [Thu, 25 Jun 2015 05:19:55 +0000 (07:19 +0200)]
info: Belatedly update for change of stop prefix to '!'

Commit eb1512d (v4.3.6) added the '=' if stopped before efficiency.
Commit 016249c (v4.3.6) changed it to '!' without updating info ship,
plane, land, nuke.

Reported-by: Harald Katzer
Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoinfo/version: Update example to current output
Markus Armbruster [Sun, 21 Jun 2015 07:10:33 +0000 (09:10 +0200)]
info/version: Update example to current output

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoinfo: Fix option NOMOBCOST misinformation
Markus Armbruster [Sun, 21 Jun 2015 06:59:01 +0000 (08:59 +0200)]
info: Fix option NOMOBCOST misinformation

The cost of firing naval guns is 15 mobility with option NOMOBCOST
disabled.  Mobility.t is correct.

Fix Options.t not to claim submarines pay half the sector movement
cost when NOMOBCOST is enabled.

Fix fire.t not to claim ships pay half the sector movement cost when
NOMOBCOST is disabled.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoinfo/Options: Nicer markup, more consistent format
Markus Armbruster [Sun, 21 Jun 2015 06:32:47 +0000 (08:32 +0200)]
info/Options: Nicer markup, more consistent format

Don't list options separately for major server versions.  It's only of
historical interest, which "info History" satisfies.

Make it a list (.L) instead of preformatted text (.nf).

Fix up so the option explanations are full sentences, starting with a
capital letter and ending with a period.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoinfo/Options: Belatedly remove SAIL
Markus Armbruster [Sun, 21 Jun 2015 06:05:24 +0000 (08:05 +0200)]
info/Options: Belatedly remove SAIL

Missed in commit dc73207.

Reported-by: Harald Katzer
Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agolwp: Rename LwpSigCatched to LwpSigCaught
Markus Armbruster [Sun, 14 Jun 2015 09:44:19 +0000 (11:44 +0200)]
lwp: Rename LwpSigCatched to LwpSigCaught

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agodocs/coding: Explain function/struct/union comment conventions
Markus Armbruster [Sun, 14 Jun 2015 09:39:07 +0000 (11:39 +0200)]
docs/coding: Explain function/struct/union comment conventions

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoFix and clean up some comments
Markus Armbruster [Sun, 14 Jun 2015 09:36:40 +0000 (11:36 +0200)]
Fix and clean up some comments

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoChange comment style to use @foo rather than FOO
Markus Armbruster [Sun, 14 Jun 2015 08:33:43 +0000 (10:33 +0200)]
Change comment style to use @foo rather than FOO

... when referring to a function's parameter or a struct/union's
member.

The idea of using FOO comes from the GNU coding standards:

    The comment on a function is much clearer if you use the argument
    names to speak about the argument values.  The variable name
    itself should be lower case, but write it in upper case when you
    are speaking about the value rather than the variable itself.
    Thus, "the inode number NODE_NUM" rather than "an inode".

Upcasing names is problematic for a case-sensitive language like C,
because it can create ambiguity.  Moreover, it's too much shouting for
my taste.

GTK-Doc's convention to prefix the identifier with @ makes references
to variables stand out nicely.  The rest of the GTK-Doc conventions
make no sense for us, however.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoconfigure: Suppress Clang warning on carg()
Markus Armbruster [Sat, 6 Jun 2015 17:44:46 +0000 (19:44 +0200)]
configure: Suppress Clang warning on carg()

Renaming carg() would be smarter, but I'd rather do that as part of a
consistent renaming of all command functions, and I'm not up to that
right now.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoconfigure: Enable -Wformat-security
Markus Armbruster [Sat, 6 Jun 2015 16:09:02 +0000 (18:09 +0200)]
configure: Enable -Wformat-security

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agopr: Make prdate() more obviously correct
Markus Armbruster [Sat, 6 Jun 2015 16:01:42 +0000 (18:01 +0200)]
pr: Make prdate() more obviously correct

Using ctime() as pr()'s first argument is safe, because its value
never contains '%'.  Clean it up anyway, so we can enable
-Wformat-security.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
6 years agoconfigure: Enable more gcc warnings
Markus Armbruster [Sat, 6 Jun 2015 15:46:56 +0000 (17:46 +0200)]
configure: Enable more gcc warnings

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoconfigure: Test compiler flags
Markus Armbruster [Sat, 6 Jun 2015 13:53:00 +0000 (15:53 +0200)]
configure: Test compiler flags

When AC_PROG_CC detects GCC, Make.mk adds a bunch of flags to CFLAGS.
Works only for flags that any version of gcc in use accepts.

Instead, make configure add the flags that actually work to CFLAGS.
This will let us add flags that work only for some compilers.

The new autoconf macros are from autoconf-archive v2015.02.24.

Unfortunately, AX_APPEND_COMPILE_FLAGS doesn't work reliably for
-Wno-*: gcc complains about unknown -Wno-foo only when other
diagnostics are being produced.  Test -Wfoo instead of -Wno-foo, and
rename to MY_APPEND_COMPILE_FLAGS.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoconfigure: Drop unused AC_C_CHAR_UNSIGNED test
Markus Armbruster [Sat, 6 Jun 2015 12:37:39 +0000 (14:37 +0200)]
configure: Drop unused AC_C_CHAR_UNSIGNED test

Unused since commit 0c8c169. v4.3.3.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoBump version to 4.3.34
Markus Armbruster [Thu, 4 Jun 2015 08:52:11 +0000 (10:52 +0200)]
Bump version to 4.3.34

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoUpdate change log timestamp for 4.3.33 v4.3.33
Markus Armbruster [Wed, 20 May 2015 18:20:40 +0000 (20:20 +0200)]
Update change log timestamp for 4.3.33

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoRevert "commands: Always put ship or land unit before retreating it"
Markus Armbruster [Sun, 17 May 2015 16:11:20 +0000 (18:11 +0200)]
Revert "commands: Always put ship or land unit before retreating it"

This reverts commit f4d8d64bb34dee75cb2485c130e125fdd3980c18.

Breaks retreat after ship got sunk by bombs or missile.

ship_bomb() and launch_missile() pass .shp_own to retreat_ship().
Wrong after putship(), because putship() resets the owner when the
ship got sunk.  retreat_ship() then oopses and fails to retreat the
surviving members of the group.

Other callers save the owner before putting the ship, and pass that.
We could change these two to do the same.  But since we're trying to
get a release out, simply revert the broken commit instead.

7 years agoTidy up some comments
Markus Armbruster [Sun, 17 May 2015 16:08:45 +0000 (18:08 +0200)]
Tidy up some comments

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoUpdate change log again for 4.3.33
Markus Armbruster [Thu, 14 May 2015 07:46:13 +0000 (09:46 +0200)]
Update change log again for 4.3.33

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoImprove portability to really outmoded compilers
Markus Armbruster [Thu, 14 May 2015 07:27:05 +0000 (09:27 +0200)]
Improve portability to really outmoded compilers

A few modernisms have crept in:

* Trailing comma in enum definition (commit 71320ed, v4.3.10)

* // comment (commit 265e71e, v4.3.11)

* <stdint.h> (commit 9102ecc, v4.3.31)

  MSC actually chokes on this one.

Avoid them.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoxundump: Pass unsigned * for scanf directive %3o, not int *
Markus Armbruster [Thu, 14 May 2015 07:03:27 +0000 (09:03 +0200)]
xundump: Pass unsigned * for scanf directive %3o, not int *

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agordsched: Pass unsigned * for scanf directive %u, not int *
Markus Armbruster [Thu, 14 May 2015 06:57:52 +0000 (08:57 +0200)]
rdsched: Pass unsigned * for scanf directive %u, not int *

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoinfo/History: Cover removal of Autonav, SAIL and TREATIES
Markus Armbruster [Sun, 5 Apr 2015 09:53:07 +0000 (11:53 +0200)]
info/History: Cover removal of Autonav, SAIL and TREATIES

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agow32/strptime: Fix for 64 bit pointers
Markus Armbruster [Sun, 5 Apr 2015 07:45:11 +0000 (09:45 +0200)]
w32/strptime: Fix for 64 bit pointers

The __UNCONST() stolen from NetBSD assumes unsigned long can hold a
pointer.  Not true with Win64's LLP64 data model.  There, we cast the
64 bit pointer to 32 bits and back.  Works only because Windows puts
the stack at a very low address, and the casts don't actually change
the pointer.

Dumb it down to a straight cast to void * for safety.

Thanks to Harald Katzer and Ron Koenderink for their help figuring out
the bug's impact.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoUpdate change log again for 4.3.33
Markus Armbruster [Fri, 13 Mar 2015 06:07:03 +0000 (07:07 +0100)]
Update change log again for 4.3.33

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoINSTALL: Refresh from automake 1.13
Markus Armbruster [Sun, 8 Mar 2015 17:23:33 +0000 (18:23 +0100)]
INSTALL: Refresh from automake 1.13

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoman: Get rid of unwanted space in SYNOPSIS
Markus Armbruster [Sun, 8 Mar 2015 16:48:17 +0000 (17:48 +0100)]
man: Get rid of unwanted space in SYNOPSIS

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoclient: New option -r for restricted mode
Marisa Giancarla [Sat, 7 Dec 2013 15:43:13 +0000 (16:43 +0100)]
client: New option -r for restricted mode

Redirections and the execute command let the user read and write files
and run programs on the local system.

Restricted mode prevents such access.  This is useful when you want to
grant somebody access to just Empire, but not to the host system's
user account that runs the client.

Signed-off-by: Marisa Giancarla <fstltna@me.com>
Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agotorpedo mission: Print rounded, not truncated hit chance
Markus Armbruster [Sun, 8 Mar 2015 13:58:04 +0000 (14:58 +0100)]
torpedo mission: Print rounded, not truncated hit chance

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agobuild: Fix inexact calculation of required materials
Markus Armbruster [Sun, 8 Mar 2015 13:06:02 +0000 (14:06 +0100)]
build: Fix inexact calculation of required materials

sector_can_build() computes mat[i] * (effic / 100.0).  The division is
inexact.  The result gets randomly rounded, so errors are vanishingly
unlikely to screw up material consumption.

However, we require the amount rounded up to be present since commit
1227d2c.  Errors *can* screw that up.  Fix by avoiding inexact
computation for that part.

We should probably review rounding of inexact values in general.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agotests/README: Cover info.ps and document .NA use for commands
Markus Armbruster [Sun, 8 Mar 2015 11:44:02 +0000 (12:44 +0100)]
tests/README: Cover info.ps and document .NA use for commands

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agotests/README: Update for replacement of info/checklist.pl
Markus Armbruster [Sun, 8 Mar 2015 11:42:00 +0000 (12:42 +0100)]
tests/README: Update for replacement of info/checklist.pl

Replaced by tests/info-test in commit 90eaf9d.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agobuild-aux: Refresh from automake 1.13
Markus Armbruster [Thu, 5 Mar 2015 20:23:46 +0000 (21:23 +0100)]
build-aux: Refresh from automake 1.13

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agom4: Refresh from autoconf-archive v2015.02.24
Markus Armbruster [Wed, 4 Mar 2015 19:07:00 +0000 (20:07 +0100)]
m4: Refresh from autoconf-archive v2015.02.24

commit dd946914edb9da25e2d8c8b2186fab5d8103f3d4
Author: Daniel Richard G <skunk@iSKUNK.ORG>
Date:   Thu Oct 3 18:56:24 2013 +0200

    AX_PTHREAD: add support for Clang

    Further details can be found at <http://savannah.gnu.org/patch/?8186>.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agotests: Rewrite the previous commit's sed_i helper
Markus Armbruster [Wed, 4 Mar 2015 07:15:13 +0000 (08:15 +0100)]
tests: Rewrite the previous commit's sed_i helper

Last night's version looks unnecessarily ugly in the morning.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agotests: Avoid non-portable sed -i
Markus Armbruster [Tue, 3 Mar 2015 19:45:03 +0000 (20:45 +0100)]
tests: Avoid non-portable sed -i

Crept in in commit 7c3186f.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agotests: Fix for builds outside git-controlled source tree
Markus Armbruster [Tue, 3 Mar 2015 19:09:50 +0000 (20:09 +0100)]
tests: Fix for builds outside git-controlled source tree

We run "git ls-files" in the build tree.  Doesn't work when the source
directory isn't a git repository, or the build directory is outside
the source directory.  Broken in commit 71cb2d8.

Find source files like Make.mk does: if the source tree is a git
repository, use git ls-files, else use sources.mk.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoBelatedly bump version to 4.3.33
Markus Armbruster [Mon, 2 Mar 2015 20:16:27 +0000 (21:16 +0100)]
Belatedly bump version to 4.3.33

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agobridgefall: Clean up literal 20 to SCT_MINEFF
Markus Armbruster [Sun, 1 Mar 2015 17:02:14 +0000 (18:02 +0100)]
bridgefall: Clean up literal 20 to SCT_MINEFF

Messed up in commit 4824648.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoUpdate change log for 4.3.33
Markus Armbruster [Sun, 1 Mar 2015 10:14:00 +0000 (11:14 +0100)]
Update change log for 4.3.33

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoadd: Drop syntax deprecated in 4.3.29
Markus Armbruster [Sun, 1 Mar 2015 09:02:25 +0000 (10:02 +0100)]
add: Drop syntax deprecated in 4.3.29

Deprecated in commit 373f20a: argument "new".

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agosubs/move: Drop syntax deprecated in 4.3.27
Markus Armbruster [Sun, 1 Mar 2015 08:58:55 +0000 (09:58 +0100)]
subs/move: Drop syntax deprecated in 4.3.27

Deprecated in commit 28d4847: no space between 'm' and its first
argument.  Affects explore, move, test, transport.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agomaps: Drop bmap flags syntax deprecated in 4.3.27
Markus Armbruster [Sun, 1 Mar 2015 08:47:19 +0000 (09:47 +0100)]
maps: Drop bmap flags syntax deprecated in 4.3.27

Deprecated in commit a00f9e2: 'r' with flags, and bad flags after 't'.
Affects flags argument of bmap, sbmap, pbmap, lbmap, nbmap, and
navigate and march sub-command 'B'.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoinfo/fire: Drop misinformation on damage varying with gun size
Markus Armbruster [Sun, 22 Feb 2015 09:55:40 +0000 (10:55 +0100)]
info/fire: Drop misinformation on damage varying with gun size

Goes all the way back to Empire 1, and as far as I can tell was
misleading there already.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoinfo/fire info/torpedo: Purge references to option MULTIFIRE
Markus Armbruster [Sun, 22 Feb 2015 09:49:58 +0000 (10:49 +0100)]
info/fire info/torpedo: Purge references to option MULTIFIRE

MULTIFIRE became non-optional in Chainsaw 3, more than two decades
ago.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoemp_config: Don't monkey-patch WORLD_X to be even
Markus Armbruster [Sun, 22 Feb 2015 07:04:41 +0000 (08:04 +0100)]
emp_config: Don't monkey-patch WORLD_X to be even

emp_config() silently truncates WORLD_X to even.  Drop that.  We could
flag odd WORLD_X as error, but we don't validate the other
configuration values, so why this one?  Instead document it needs to
be even.  WORLD_Y, too.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoef_verify: Polish product error message
Markus Armbruster [Sat, 21 Feb 2015 18:07:44 +0000 (19:07 +0100)]
ef_verify: Polish product error message

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoinit ef_verify: Don't monkey-patch capability VTOL, require it
Markus Armbruster [Sat, 21 Feb 2015 17:56:32 +0000 (18:56 +0100)]
init ef_verify: Don't monkey-patch capability VTOL, require it

A plane with capability missile must have capability VTOL.  When it's
missing, global_init() silently adds it.

Drop that.  Check for it in ef_verify_config() instead.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoinit ef_verify: Don't monkey-patch capability miss, require it
Markus Armbruster [Sat, 21 Feb 2015 17:19:39 +0000 (18:19 +0100)]
init ef_verify: Don't monkey-patch capability miss, require it

A ship with non-zero nplanes must have capability plane or miss.  When
one doesn't, global_init() silently adds capability miss.

Drop that.  Check for it in ef_verify_config() instead.  Fix up
ship.config accordingly.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agoef_verify: Require ship glim != 0 with capability dchrg
Markus Armbruster [Sat, 21 Feb 2015 17:12:07 +0000 (18:12 +0100)]
ef_verify: Require ship glim != 0 with capability dchrg

Because the fire command doesn't work with zero glim (commit a109de9).

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agotorpedo: Suppress bulletin when player torpedoes his own ship
Markus Armbruster [Sun, 15 Feb 2015 14:22:01 +0000 (15:22 +0100)]
torpedo: Suppress bulletin when player torpedoes his own ship

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agotorpedo: Print "Starting our attack run" regardless of target
Markus Armbruster [Sun, 15 Feb 2015 14:06:14 +0000 (15:06 +0100)]
torpedo: Print "Starting our attack run" regardless of target

Instead of printing it only when the target owned by somebody else.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
7 years agotorpedo: Let torpedo hit land only when target is in range
Markus Armbruster [Sun, 15 Feb 2015 13:25:39 +0000 (14:25 +0100)]
torpedo: Let torpedo hit land only when target is in range

Telling the player his torpedo "slams into land" can give a clue on
the direction to the target.  No good when the target is out of range,
because we shouldn't tell the player more than that then.

Screwed up in 4.2.2.  Fix by checking range before line of sight.

Signed-off-by: Markus Armbruster <armbru@pond.sub.org>