So you don't have to micromanage workers to maximize useful work.
The previous commit made the problem a bit worse. If you had a few
workers too many before, you perhaps produced an extra unit. Now, you
get to keep the extra work instead. Useless, unless it rolls over.
produce() limits production to how many units the workers can produce,
rounding randomly. It charges work for the units actually produced,
discarding fractions.
If you get lucky with the random rounding, you may get a bit of extra
work done for free. Else, you get to keep the unused work, and may
even be undercharged a tiny bit of work. Has always been that way.
The production command assumes the random rounding rounds up if and
only if the probability to do so is at least 50%. Thus, it's
frequently off by one for sectors producing at their worker limit.
The budget command runs the update code, and is therefore also off by
one, only differently.
Rather annoying for tech and research centers, where a single unit
matters. A tech center with full civilian population can produce 37.5
units in 60 etus. Given enough materials, it'll fluctuate between 37
and 38. Production consistently predicts 38, and budget randomly
predicts either 37 or 38. Both are off by one half the time.
Fix this as follows: limit production to the amount the workers can
produce (no rounding). Work becomes a hard limit, not subject to
random fluctuations. Randomly round the work charged for actual
production. On average, this charges exactly the work that's used.
More importantly, production and budget now predict how much gets
produced more accurately. They're still not exact, as the amount of
work available for production remains slightly random.
This also "fixes" the smoke test on a i686 Debian 6 box for me. The
root problem is that floating-point subexpressions may either be
computed in double precision or extended precision. Different
machines (or different compilers, or even different compiler flags)
may use different precision, and get different results.
Example: producing 108 units at one work per unit, sector p.e. 0.4
needs to charge 108 / 0.4 work. Computed in double precision, this
gets rounded to 270.0, then truncated to 270. In 80 bit extended
precision, it gets rounded to 269.999999999, then truncated to 269.
With random rounding instead of truncation, the probability for a
different result is vanishingly small. However, this commit
introduces truncation in another place. It just happens not to mess
up the smoke test there. I doubt this is the last time this kind of
problem upsets the smoke test.
We can make actual = roundavg(material_consume * prodeff) products.
When we reduce actual, we have to reduce material_consume, too. Code
does that like this:
material_consume = roundavg(actual' * material_consume / actual)
Double rounding. Do this instead:
material_consume = roundavg(actual' / prodeff)
Broken in commit 3a7d7fa, which enlarged struct natstr member
nat_hostaddr[] from 32 to 46 characters, but neglected to update the
ca_len in nat_ca[]. Consequently, the address is truncated in xdump.
Can also break country * ?ip=... and such, but that's exotic.
emp_server and empdump refuse to start on most big endian hosts,
because ef_verify_config() chokes on mdchr_ca[]:
Config meta uid 0 field type: value 0 is not in symbol table meta-type
Config meta uid 1 field type: value 0 is not in symbol table meta-type
Config meta uid 2 field type: value 0 is not in symbol table meta-type
Config meta uid 3 field type: value 0 is not in symbol table meta-type
Config meta uid 4 field type: value 0 is not in symbol table meta-type
Broken in commit 06a0036 (v4.3.12), which changed struct castr member
ca_type from packed_nsc_type (typedef'ed to char) to enum nsc_type,
but neglected to update the ca_type in mdchr_ca[].
On little endian hosts, the selector reads the least significant byte,
with sign extension. Happens to work, because the type values are all
sufficiently small integers.
On big endian hosts, the selector reads the most signiciant byte.
which is always zero (NSC_NOTYPE). Makes ef_verify_config() fail.
Except when sizeof(enum nsc_notype) == 1. Then selector type works
fine, and ef_verify_config() succeeds, but we run into the next
problem: the same commit also changed member ca_flags from nsc_flags
(typedef'ed to unsigned char) to int without updating the ca_type in
mdchr_ca[]. This breaks "only" xdump meta column flags.
v4.3.12 was released in April 2008. Either nobody has tried to run a
game on a big endian host since, or all who did gave up quietly,
without reporting the problem.
We clearly need to test on a wider range of machines.
Broken in commit 14ea670 (v4.3.8), which changed struct trdstr member
trd_type from char to short, but neglected to update the ca_type in
trade_ca[].
On little endian hosts, the selector reads the least significant byte,
with sign extension. Happens to work, because the type values are all
sufficiently small integers.
On big endian hosts, the selector reads the most signiciant byte,
which is always zero (EF_SECTOR). Messes up xdump trade badly.
Broken in commit 09248d0 (v4.3.8), which changed struct loststr member
lost_type from char to short, but neglected to update the ca_type in
lost_ca[].
On little endian hosts, the selector reads the least significant byte,
with sign extension. Happens to work, because the type values are all
sufficiently small integers.
On big endian hosts, the selector reads the most signiciant byte,
which is always zero (EF_SECTOR). Messes up xdump lost badly. Also
breaks lost * ?type=..., but that's exotic.
"Unusually long" topics are marked with a "!" in subject indexes.
This should use the line count of the formatted page, but that's too
much trouble, so commit 4c0b4c0 (v4.3.27) approximated it by "source
file has more than 9999 bytes". Change that to "source file has more
than 300 lines".
Since subjects were added in Empire 2, we've always picked them up
from .SA requests. If you mistype a subject there, you get a "is a
NEW subject" warning, and incorrect subject pages. When building a
pristine tree, you get bogus "is a NEW subject" warnings for all
subjects. If you somehow delete the generated subjects.mk, but not
the generated subject files, the build breaks.
Declare subjects in Make variable subjects. Drop generated makefile
subject.mk.
Treat unknown topics in .SA arguments as errors. This replaces the
"$subj is a NEW subject" warning.
Treat subjects without member pages as errors. This replaces the "The
subject $subj has been removed" warning.
Safer and simpler.
We used to do all the info indexing work in info.pl: find subjects,
create subjects.mk (to tell make the list of subjects), the subject
pages, and TOP.t. Worked, but touching an info page triggered a full
rebuild of all subject pages and TOP.t.
Commit 2f14064 (v4.3.0) tried to avoid that by splitting info.pl into
findsubj.pl, mksubj.pl, mktop.pl. findsubj.pl puts not just the
subjects into subjects.mk, but also make rules for the subject pages,
to guide their remaking. mksubj.pl creates a single subject page.
mktop.pl creates TOP.t.
Unfortunately, this doesn't work so well. Since subjects.mk doesn't
exist in a virgin tree, we use -include. Unwanted consequence:
findsubj.pl failure doesn't stop make. Moreover, the complex make
machinery breaks down when info sources get removed or subjects get
dropped.
Go back to the old method, except keep mktop.pl separate, as that part
works just fine, and use simpler make rules. mksubj.pl now creates
subjects.mk and all subject pages, like info.pl did.
This effectively reverts most of commit 2f14064. I'll address the
excessive rebuilding of subject pages in a different way shortly.
Remaking config.h and config.h.in updates the target only when its
contents actually changes. This is important, because after updating
config.h we need to recompile everything.
The make rules to do that are straight from the Autoconf manual. But
they don't work: the rules that connect config.h and config.h.in to
stamp-h and stamp-h.in don't have recipes. Since make doesn't have an
implicit rule either, it concludes that the target remains unchanged.
It still updates the prerequisites. The recipe for updating the stamp
files then change the the targetes behind make's back. Make misses
the change of config.h and/or config.h.in, and we get an incomplete
rebuild.
The rules need empty recipes instead. This Autoconf manual was fixed
accordingly in autoconf.git commit 6b42b38.
Mark obsolete pages with '+' in subject pages. Drop the separate
"Obsolete" subject page: move "info Innards" to subject "Server", and
"info update" to "Updates" (where it came from in commit a5764534,
v4.3.10).
For instance, use "127.0.0.1" for IPv4 loopback instead of
"::ffff:127.0.0.1".
Simplifies use of econfig key privip: plain dotted decimal now just
works regardless of IPv6 use, no need to add the IPv4-mapped form.
Also affects how addresses are logged and shown to players, and nation
selector ip. Nicer that way.
Systems using GNU libc such as Linux are frequently configured in a
way getaddrinfo(NULL, ...) put the IPv4 wildcard "0.0.0.0" *before*
the IPv6 wildcard "::" in the result. Because of that, listen_addr ""
listens only on all IPv4 addresses. Workaround: listen_addr "::".
Document it in listen_addr's doc string.
OpenBSD refuses to implement IPV6_V6ONLY, in violation of RFC 3493.
RFC 4038 frowningly recognizes this practice. The only way to bind
both IPv4 and IPv4 there is two separate sockets. Requires more
surgery than I can do now.
Since we can't have both IPv6 and IPv6 on OpenBSD with our single
socket, prefer IPv4, but if that doesn't work, do IPv6.
To prefer IPv6 instead, put 'listen_addr "::"' into econfig. Document
that in listen_addr's doc string.
We rely on AF_INET6 wildcard bind() binding the AF_INET port, too,
i.e. IPV6_V6ONLY off. This should be the default according to RFC
3493 section 5.3, but isn't on Windows and BSD. RFC 4038 recognizes
this fact in section 4.2.
When IPV6_V6ONLY is on, an AF_INET6 wildcard bind only accepts
connections from IPv6 addresses. Thus, IPv4 doesn't work when
getaddrinfo() returns an AF_INET6 address first (which it should do
when the system has an IPv6 address configured).
Switch off IPV6_V6ONLY explicitly instead of relying on the default.
This makes IPv6 work on systems where IPV6_V6ONLY is on by default,
such as Windows and BSD.
Except for OpenBSD, which does not support switching it off. To be
addressed in the next commit.
Shouldn't fail. If it fails, but bind() works, the failure doesn't
matter. If bind() fails, we can just as well report that failure
instead of setsockopt()'s.
How qsort() sorts members that compare equal is unspecified. Can
upset the smoke test. Observed under FreeBSD 8.3.
Break ties in power by comparing country numbers. Countries equal in
power are now sorted by increasing country number.
Island size is randomly chosen from the interval [1..2*is+1], with
expected value is. Use two dice to roll the size instead of one.
This makes extreme sizes much less likely.
