44 KiB
Release notes for Inform v8.4 (build 6F95)
This release was made on 25 October 2010, before Inform was open-sourced. At that time bugs were tracked by a Mantis server, with seven-digit issue numbers, and versions of Inform were identified by build codes in the form NLNN. Suggestion numbers, also large but not starting with a 0, referred to tickets at a Uservoice forum then in use.
This is part of the historical archive of release notes.
Overview
This maintenance release fixes every core Inform issue logged on the Inform bug tracker up to today, which means that it resolves around 180 issues; this change log lists most of them, but omits a few which were simple corrections of typos, and of course doesn't list the small number of issues closed as being misunderstandings. Though there are no significant new features, two additional resources have been added to the core Inform distribution: Andrew Plotkin's Quixe interpreter, allowing Glulx story files to be released as playable websites (something up to now allowed only for Z-machine story files); and Eric Eve's extension "Epistemology", which joins the built-in set for the first time.
Language
The last major release of Inform (6E59) led with an important new feature: an integrated way to publish IF to a website allowing the story file to be played in-browser. This was done by constructing the site to use the Parchment interpreter, which ships as part of the Inform distribution. However, since Parchment is an interpreter for the Z-machine VM, this meant that only projects with the format setting set to "Z-code" could be released in this way. The new release comes with not only Parchment but also Quixe, Andrew Plotkin's in-browser interpreter for the Glulx VM. In response to the instruction:
Release along with an interpreter.
Inform will automatically choose Parchment for a Z-machine project, or Quixe for a Glulx one.
The "Standard" template for websites has been updated to make in-browser play easier to style. The interpreter body is now wrapped in a div called "gameport", and the CSS stylesheet in "Standard" contains explicit layout information for the in-browser play page. This layout automatically simplifies on iPhone-sized displays; the left-hand column disappears.
"Parchment" has also been updated. There are minor bug fixes, and command input is now done with an invisible form field, so that standard line-editing conventions now all work. (Mouse selection, all the usual control keys, etc.)
The extension Epistemology by Eric Eve has been added to the built-in set in the standard Inform distribution. (The version used is version 6, identical to the one available at the Inform website since 19 June 2010.) This is a simple and stable extension, providing facilities to keep track of what the protagonist knows; we're including it in the built-in set because it is so widely used, and has become a de facto standard. Our thanks to Eric, of course.
Temporary "let" variables are now allowed to have names clashing with non-local ones; as in languages like C, these temporary definitions override the existing ones, but last only while the variable lasts. Thus:
The target is 17.
When play begins:
say "The target is [target]."
When play begins:
let the target be "brainz";
say "This target is [target]."
prints up "The target is 17.", and then "This target is brainz." This should reduce name-clash problems where it's tricky to define global variables with the same name as any local ones existing in the Standard Rules (such as, in fact, the name "target").
The rules on how temporary "let" variables are have been enforced better in this new build. This does cause some existing source text to fail. For example, in the example "Nickel and Dimed", previous builds included this:
if the total price of underpayment money is less than the cost:
let the desired bill be the cheapest money which is overpayment;
otherwise:
let the desired bill be the costliest money which is underpayment;
decide on the desired bill.
This is now rejected because "desired bill" is a name given to two different temporary values, each of which expires at the end of a branch of the "if"; so that nothing called "desired bill" still exists at the "decide on..." line. While the above source looks (and is) harmless enough, consider this:
if the total price of underpayment money is less than the cost:
let the desired bill be the cheapest money which is overpayment;
decide on the desired bill.
This is clearly unfortunate if the condition fails, as then "desired bill" is a temporary value never brought into existence at all. Past builds of Inform would have allowed the usage, giving "desired bill" its default value in that case (albeit failing badly if the kind was such that no default value existed), but this was not a good idea. So we are now enforcing the rules more strictly. The same applies to the branches of an "if X is: ...", too. The relevant passage in "Nickel and Dimed" now reads:
if the total price of underpayment money is less than the cost:
decide on the cheapest money which is overpayment;
otherwise:
decide on the costliest money which is underpayment.
The handling of the "described/undescribed" either/or property has been made more consistent, and the documentation now says this:
There is also an either/or property called "described"/"undescribed",
intended to be used only as a last resort, but which has the ability
to hide something from room descriptions. This not really hiding: the
idea is that "undescribed" should be used only for cases where some
other text already reveals the item, or where its presence is
implicit. Even then, it should only be used when the item is intended
to be taken or moved by the player at some point - if the item isn't
intended to move, it's much better to make it "scenery". (There's only
one commonly-found example - the player's own body, the "yourself", is
undescribed.)
Note that the "undescribed" property is automatically removed from
anything carried by, worn by or part of the player, even indirectly;
and that nothing on top of an "undescribed" supporter will be visible
in a room description, even if it itself is "described". (Scenery
supporters don't suffer from that restriction, which is one reason
scenery is a better option when possible.)
The rule on when punctuation in the source text divides up sentences has been slightly refined.
(a) Where a single punctuation mark other than "[" or "]" occurs in between two digits, or between a digit and a minus sign, or (in the case of full stops) between two lower-case alphanumeric characters, no division is made. This is done so that, for instance, "0.91" does not split into three words.
(b) Where the character following the punctuation mark is a slash, no division is made. This is done essentially to make most common URLs glue up as single words; otherwise "http://" would be broken up.
The clauses in an action definition ("requiring light", "applying to...", and so on) previously had to appear in a specific order; now they can appear in any order.
Implementing suggestion 1015949: RULES ON and SCENES ON are now synonymous with RULES and SCENES. (Previously OFF worked, but ON didn't, which was inconsistent with the behaviour of the ACTIONS command. This is all now consistent.)
Documentation, examples and extensions
Numerous typos have been removed from both manuals. In "Writing with Inform", description of "handling an activity" tweaked to be clearer, since explanations added during the creation of the phrase book misleadingly suggested this phrase was equivalent to checking whether the activity was happening.
Examples
-
In the "Recipe Book", a new section on "Timed Input" has been added, pointing users to the extensions capable of handling realtime.
-
"Ant-Sensitive Sunglasses" added to help explain the function of activities, and demonstrate how to hook activities into room and object descriptions.
-
"Unexamined Life" updated to include the different Test me required when running the example under Glulx.
-
"Blankness", "Walls and Noses", "Paddington", "Facts Were These", "A Day for Fresh Sushi", "Control Center": typos removed.
-
"Newbie": bug in the newbie triggering removed; example simplified, since it is no longer trying to demonstrate procedural rules.
-
"GET AXE", "Gelato", "Muppets": procedural rules stripped out of examples.
-
"IQ Test": removed explanation text that no longer applies now that procedural rules have been removed from the example.
-
"Timeless", "We": removed explanation about not using procedural rules, since these are now deprecated anyway.
-
"A Day for Fresh Sushi": minor tweak to deal with the bug that >GIVE FOOD TO FISH was accepted as a winning move even though some code was included to encourage the player to FEED FISH instead.
Extensions
-
"Locksmith": Version 11 removes a double-printing of the items on a keyring, arising because of changes to the Standard Rules for describing examined supporters.
-
"Plurality": Version 9 re-includes some phrases that were introduced (including 's/'re and more rational capitalization schemes); and also names several previously unnamed rules, and moves pronoun-noticing rules to a more appropriate place in the turn sequence.
-
"Glulx Entry Points": Version 9 adds new wrappers for glk gestalts to detect whether char input and specific types of mouse input (graphics windows vs text-grid windows) are supported, and divides up the code with section headers to make it easier to surgically replace sections of the extension.
-
And, as noted above, "Epistemology" has been added, at version 6.
Problem messages
Bug fixes
Source text and punctuation
Bug fixed (0000345) whereby "To" phrase could be defined with quoted text as part of the fixed wording, even though this isn't allowed and doesn't work. There's now a problem message to cover it.
(See also "Spacing and Printing" below, and see the notes above about word-breaking conventions.)
Headings
Bug fixed (0000231) whereby a title which contains a paragraph break (why, why, would anyone do this?) would have it appearing with HTML tags printed up as part of the title when the story file begins play.
Bug fixed (0000161) whereby sections with quotation marks in their headings can't be replaced by "in place of" headings.
Extensions
Bug fixed (0000225) whereby spacing around punctuation, when extension documentation is being pretty-printed, would sometimes be excessive.
Bug fixed (0000132) whereby extension documentation containing examples with text containing literal paragraph breaks would fail to set up the "paste source" button correctly.
Bug fixed (0000205) whereby extension documentation containing examples with text containing a (- ... -) I6 escape sequence would fail to set up the "paste source" button correctly.
Bug fixed (0000167) whereby tooltips (i.e. explanation text appearing above the mouse arrow when it hovers) over a link icon to extension documentation would name the extension and author with Each Word Capitalisation, instead of respecting the capitalisation actually used by the extension, e.g., "CrAzYcApS by UNUSUALLY caPitAlized namE" as "Crazycaps By Unusually Capitalized Name".
Bug fixed (0000230) whereby the wording used on the warning for an incorrectly installed extension implied that extensions aren't allowed to have filename extensions - whereas they now can, and indeed should, end in ".i7x".
Bug fixed (0000217) whereby extensions with prodigiously large quantities of documentation (say 500,000 words or more) would, in some circumstances, crash Inform when used.
Assertions and creations
Bug fixed (0000339) whereby an internal error rather than a problem message would result from trying to set properties of a "kind", like so:
The printed name of a kind is usually "xyzzy."
Bug fixed (0000340) whereby a rather vague problem message would be issued in response to (also rather vague!) sentences such as:
A kind has a number called X.
Bug fixed (0000341) whereby attempts to create a "kind of kind" do not lead to problem messages.
Bug fixed (0000343a) whereby attempts to create a "kind of every X" do not lead to problem messages.
Bug fixed (0000343b) whereby attempts to create a kind uncertainly do not lead to problem messages.
Bug fixed (0000254) whereby creations of rulebooks no longer lead to misreadings of "it". For example,
The Test rules is a rulebook.
It has a number called the test number.
now produces a problem message; previously, "it" would have been taken to refer to something prior to the first sentence here, with disastrous consequences.
Bug fixed (0000109) whereby no problem message is issued if the source tries to create an object using an adjective which is an antonym; e.g.,
Definition: A device is complex rather than simple if its CPU power is
50 or more.
A simple device called a netbook is in the lab.
was previously accepted but created the netbook as a complex device; it now produces a problem message, since Inform doesn't know what CPU power to assign to the netbook.
Bug fixed (0000138) whereby an unhelpful problem message was produced in reply to ambiguities such as this one:
The kitchen is a room. The player carries no tea.
Every turn: if the player carries no tea, say "Still no tea."
where an object called "no tea" is created, but this clashes against "no" as a determiner. A somewhat better problem message is now produced.
Bug fixed (0000195) whereby a nameless room passively created could not always be referred to implicitly by "here", as in this example:
There is a dark room. The player is here.
Bug fixed (0000139) whereby Inform silently ignored assertions about the value of a variable which aliases an I6 variable, as here:
Foo is a number that varies. The foo variable translates into I6 as "foobar".
Foo is usually 5.
Inform now issues a problem message. (It's impossible because Inform 7 doesn't control this variable, and can't create or initialise it.)
Bug fixed (0000276) whereby names could clash against local variable names; see "minor new features" above.
Bug fixed (0000174) whereby sounds, figures and external files could be created implicitly without using the special syntax provided for this (but the results wouldn't always work); for example
A tag is a kind of value.
Tagging relates various tags to various figure names. The verb to tag
(he tags, they tag) implies the tagging relation.
Outdoors is a tag. Outdoors tags Figure of Forest.
would cause Inform to create the figure, since it knows from the relation that this is what "Figure of Forest" must be, but the result is unhappy because there's no way to know the filename for the image. Inform now issues an explanatory problem message for cases like this.
Bug fixed (0000218) whereby contradiction problems arising from assemblies were not being reported very clearly. It takes quite a while to realise what is wrong with the following, for instance:
There is a room.
A hairstyle is a kind of thing.
A hairspray is a kind of thing.
Every hairstyle incorporates a hairspray.
Every hairspray incorporates a propellant.
There is a hairstyle.
There is a hairstyle.
The problem here is that there is only a single "propellant" object, yet Inform tries to incorporate it in both of the hairsprays created by the hairstyles; a contradiction since it can't be in two places at once. What has really happened, of course, is that the source forgot to say "A propellant is a kind of thing". Inform previously reported this, correctly but rather unhelpfully, as a contradiction between "There is a hairstyle" and "There is a hairstyle"; it now diagnoses the fault much better.
Bug fixed (0000209) whereby an unhelpful problem message was produced in response to:
A hairstyle is a kind of thing. A hairspray is part of every hairstyle.
where the use of "every" on this side of the relationship implicitly puts the same item in an unlimited number of places; a better objection is now made to this. (Compare the related 0000218.)
Bug fixed (0000134) whereby an unhelpful problem message would result from:
Instead of exiting when the location is lounge and when Rob is in location
where the actual problem is the spurious second usage of "when", which is now explained.
Bug fixed (0000333) whereby assertions such as
Foo is a number that varies. Foo is never 5.
would apparently be allowed, but have no practical effect except to prevent "foo" from being properly created. "Always", "seldom" and "never" are now prohibited in such sentences. ("Usually" remains allowed, so that extensions can set default values for variables which can be overridden by the main source text.)
Bug fixed (0000332) whereby contradictory assertions about the initial value of a variable produce a problem message about contradictory values of a property, confusingly. (Internally, initial values of variables are stored as the values of a property called "initial value" which belongs to variables, y'see. But this is purely an implementation convenience.) The problem message has been clarified.
Model world
Bug fixed (0000202) whereby inventory listings could under rare circumstances print incorrectly with translation extensions such as Spanish by Sebastian Arg.
Bug fixed (0000278) whereby the player's holdall, if in use, would sometimes be called "nothing" if it was being used in a turn when it hadn't been held at the start of the turn.
Properties
Bug fixed (0000329) whereby a property created both explicitly and as a column in a table used to create objects would have the kind implied by the table column, even if that contradicted what the explicit declaration said; e.g.,
A foo is a kind of thing. A foo has a number called the fooness.
Some foos are defined by the Table of Goodies.
Table of Goodies
foo fooness
Foo 1 "very foo indeed"
would not detect that "fooness" seems to be a number and a text at the same time. Inform now reports this with a problem message.
Bug fixed (0000287) whereby attempts to use properties which don't exist were not always picked up run-time problem messages. For instance, this passed without problems:
Home is a room. A direction has a number called the compass heading.
When play begins: say the compass heading of Home.
Bug fixed (0000124) whereby defining a property of an individual non-object value did not always work; for example,
Pool Game is a scene. Pool Game has a number called target score.
would leave a "target score" property which produced spurious run-time problem messages when used.
Relations
Bug fixed (0000160) whereby phrases like "list of K that R relates", where K is a kind and R a relation, did not work with statically defined relations.
Bug fixed (0000169) whereby dynamic relations would be corrupted in some circumstances if they grew beyond a certain (actually quite small) size.
Bug fixed (0000247) whereby testing some asymmetric relations with verbs would produce spurious I6 errors; for example,
A language is a kind of value. The languages are English and Perplexish.
Speaking relates one person to various languages. The verb to speak (he
speaks, they speak, he spoke, it is spoken, he is speaking) implies the
speaking relation.
When play begins: if Zora speaks English, say "Huzzah!"
is clearly correct source text, but would compile incorrect code to test "Zora speaks English".
Bug fixed (0000103) whereby lists created by relationships would sometimes have the wrong kinds, resulting in odd failures at run-time. For example:
Partnership relates various lists of numbers to various lists of numbers.
The verb to belong with (he belongs with, they belong with) implies the
partnership relation.
{3} belongs with {7}.
would not correctly initialise the relationship, so that "if {3} belongs with {7}" would then fail at run-time.
Bug fixed (0000080) whereby properties and variables with unsafe kinds could be created when their values were relations; for example,
Foo is a relation that varies.
A thing has a relation of values to values called bar.
are both unsafe because the relationships are too vaguely defined to be safe in usage. ("Foo is a relation of numbers that varies" would be fine.) This now results in a problem message.
Bug fixed (0000096) whereby dynamic relations of block values would be incorrectly tested at run-time, with e.g.
Partnership relates various indexed texts to various indexed texts.
When play begins: now the partnership relation relates "apples" to "bananas".
producing spurious "deep copy failed" messages.
Bug fixed (0000095) whereby valency changes to newly created relations were not taking effect for one-to-one relations (or equivalence relations), so that they sometimes came out as asymmetric when they should have been symmetric.
Bug fixed (0000185) whereby multiple-word verbs wouldn't always be conjugated correctly; for example,
Liking relates various people to various things.
The verb to really like (he really likes, they really like, he really
liked, it is really liked, he is really liking) implies the liking relation.
Definition: a thing is popular if it is really liked by all people.
was rejected because Inform didn't recognise "it is really liked" in the definition. This now works as intended.
Bug fixed (0000241) whereby relationships to "nothing" produced crashes in some cases instead of problem messages; for example,
Requirement relates various things to various things. The verb to require
(he requires, they require, he required, it is required, he is requiring)
implies the requirement relation.
The sufficiency is a thing. It requires nothing.
Actions
Bug fixed (0000303) whereby "undescribed" caused various infelicities: see "minor new features" above.
Bug fixed (0000321) whereby describing an action could fail if it included "that", since there would then be an ambiguity with a relative clause:
Answering someone that is unmaidenly behavior.
This used to cause a problem message, but is now accepted.
Bug fixed (0000177) whereby the use of action variables in the past tense caused "unavailable" run-time errors. This now causes a compile-time problem message. (It's wrong because action variables have only temporary existence; in the past, they were not alive to have a value.)
Bug fixed (0000193) whereby actions applying to lists (or other pointer values) could not have constant stored actions created; for example,
Extrapolating is an action applying to one list of numbers.
The best idea yet is a stored action which varies. The best idea yet is
the action of extrapolating {1, 2, 3}.
(This is highly dubious since such an action can never arise from a typed command, but we haven't actually prohibited it. If such actions cause further trouble, we may simply ban them.)
Bug fixed (0000216) whereby use of UNDO on the opening turn produced the wrong complaint.
Bug report (0000286) wasn't strictly a bug, but led to the syntax for creating actions being improved: see "minor new features" above.
Activities
Bug fixed (0000289) whereby an unhelpful problem message was given in reply to
A colour is a kind of value. The colours are red, green, blue.
Painting is an activity on colours.
when the only actual problem is that it should have been "painting something".
Rules and rulebooks
Bug fixed (0000354) whereby a rule with no name whose premiss included literal text which contained a backslash character would cause I6 errors.
Bug fixed (0000336) whereby the implied owner of a property mentioned in a rule would not always be the noun of the action to which the rule applied, depending on the sequence of prior rules.
Bug fixed (0000157) whereby use of constant values of certain kinds would not work if they were needed to determine a condition which involves an implicit search, for example, in the following search through all people implied by the word "someone":
Gambling relates various people to various lists of numbers. The verb to
bet on (he bets on, they bet on) implies the gambling relation.
Every turn when someone bets on {1, 2, 3, 4, 5}: say "Just checking."
Bug fixed (0000306) whereby rules such as this one, explicitly mentioning "an actor":
Report an actor painting when tint is red:
would fail because Inform wasn't reading "when" correctly.
Bug fixed (0000322) whereby rule-sorting did not detect that a "when" condition with two or more clauses divided by "and" was more specific than one with fewer clauses.
Bug fixed (0000125) whereby rule-sorting did not detect that a specific action like "examining the apple tree" was more specific than "doing something to the apple tree".
Bug fixed (0000176) whereby rule preambles naming bare kinds (rather than qualifying them with adjectives) would fail with spurious problems. For example,
Dialing is an action applying to one number.
Carry out dialing a number: say "There's no answer."
Bug fixed (0000072) whereby rulebooks producing pointer values (indexed texts, say, or lists) would not work. For example:
Scribbling is a rulebook producing an indexed text.
When play begins: let T be the indexed text produced by the scribbling
rulebook.
would produce a spurious "deep copy failed" error at run-time.
Bug fixed (0000264) whereby the use of the word "rulebook" wasn't always consistent with "rules", e.g., creating
The alpha rules is a rulebook.
The beta rulebook is a rulebook.
would allow rules to begin "Alpha:" but not "Beta:"; this contradicted the documentation, and has been put right.
Bug fixed (0000248) whereby callings in rules for object based rulebooks did not always work.
Bug fixed (0000179) whereby "anonymously abide by R", for some rule R, would sometimes abide by it but not anonymously - the circumstances are a little complicated to go into here, so see the tracker.
Lists
Bug fixed (0000137) whereby nested list variables were not always correctly initialised; for example,
The grid is a list of lists of indexed texts which varies. The grid is
{{"foo", "bar"}}.
Bug fixed (0000293) whereby constant lists containing indexed texts which themselves were initialised with text substitutions would not work, under some eldritch circumstances: see the tracker.
Bug fixed (0000108) whereby comparing a newly-created empty list against a constant empty list would fail:
let L be a list of numbers;
if L is {}, say "Test succeeded.";
(because Inform wrongly considered these lists to have different kinds of entries).
Bug fixed (0000224) whereby constant lists including as entries constants which themselves had names including the word "with" would be misunderstood. For example,
Ducks-in-a-row is a list of figure names that varies. Ducks-in-a-row is
{Figure of Duck with Hat, Figure of Duck with Hat}.
Tables
Bug fixed (0000304) whereby nested "repeat through the table..." loops would sometimes fail, iterating incorrectly.
Bug fixed (0000188) whereby tables which begin entirely empty of rows, and where the kinds of the columns are therefore given in brackets after the column names, could not be created.
Bug fixed (0000228) whereby testing for the existence of an entry in a blank table would result in spurious "programming error" messages at run-time.
Bug fixed (0000189) whereby excessively large numbers of table entries would cause I6 errors on the Z-machine setting, rather than a problem message. (Memory is very limited in the Z-machine.)
Scenes
Bug fixed (0000294) whereby scenes were not allowed to be joined circularly end-to-end like so:
Repetition begins when Repetition ends.
Bug fixed (0000187) whereby scenes beginning based on what encloses the player would sometimes hang at run-time on the opening turn:
Importance is a scene. Importance begins when Current Place encloses the
player for the first turn.
Bug fixed (0000337) whereby an attempt to write a condition for a scene beginning or ending, for something which isn't in fact a scene, would sometimes produce an internal error rather than a problem message.
Kinds and typechecking
Bug fixed (0000346) whereby attempts to rearrange the kinds hierarchy so as to introduce a circularity, like so:
A person is a kind of animal.
(i.e., when the reverse is already true) caused Inform to hang. There's now a problem message.
Bug fixed (0000221) whereby attempts to create an unsafe variable which contains a list of table columns would be rejected with an unhelpful problem message.
Bug fixed (0000335) whereby internal errors would sometimes result from unsafe use of mismatching descriptions, e.g., using a description of things where a description of people was needed.
Bug fixed (0000222) whereby the wrong problem message would arise from a let variable being created like so:
let C be a table column;
(This is incorrect because unsafe; C can't be used unless we know the kind of the entries in the column.)
Bug fixed (0000129) whereby assignments to variables were not always checked properly at run-time for the correct kind of object, e.g.:
The place of origin is a room that varies.
The shadows are scenery.
When play begins:
let a dim spot be the shadows;
now the place of origin is a dim spot.
is unsafe because it assigns "shadows", a thing, to "place of origin", which is supposed to contain a room. This is now caught.
Bug fixed (0000302) whereby changing a text property to an indexed text value, for example, with
To rename to (Y - some indexed text):
now the printed name of location is Y.
would be allowed but then crash at run-time. There's now a problem message at compile time.
Bug fixed (0000327) whereby assigning a value of the wrong kind to a property belonging to a constant value (which is not an object) would throw an internal error instead of a problem message.
Bug fixed (0000173) whereby spurious problem messages would be produced by a description used as a value which was written with a relative clause, for example,
To disregard (D - a description of numbers): do nothing.
When play begins: disregard numbers which are even.
where "numbers which are even" would be rejected as a description of numbers, even though that's what it is.
Bug fixed (0000184a) whereby definitions about "actions", like this one:
Definition: an action is gluttonous if it is eating.
were rejected with an unhelpful problem message - the problem is that "action name" and "stored action" are kinds but "action" is not. This is now better explained.
Bug fixed (0000184b) whereby attempts to use "action" as a token in a phrase definition were not being rejected:
To disregard (act - an action): ...
(This is legal if the phrase has an I6 inline definition, and the Standard Rules uses this to define "try", for instance, but it's otherwise illegal because "action" is not a kind of value.)
Bug fixed (0000181) whereby in some circumstances "now P is V", changing the property P of some implied property-holder to a given value V, would be rejected as if it were an attempt to compare a property (as a value in its own right) with V.
Bug fixed (0000186) whereby indexed text would not work when accessed only indirectly and with a value which was originally text; for example:
To decide which number is the length of (T - an indexed text) (this is
length extraction): decide on the number of characters in T.
When play begins: showme length extraction applied to "Hello world!"
would fail with various different symptoms depending on other source text.
Bug fixed (0000331) whereby "provides" used with implicit searches would behave incorrectly for kinds of value other than kinds of object.
Phrases and functional programming
Bug fixed (0000342) whereby phrases with spurious "(called ...)" clauses in their definitions would not always produce problem messages.
Bug fixed (0000324) whereby temporary "let" variables remained in scope after the end of an "if" branch in which they were created; see "minor new features" above.
Bug fixed (0000220) whereby spurious problem messages appeared after an "if" which was malformed, complaining about the lack of an "end if".
Bug fixed (0000311) whereby a phrase definition which used too many "let" and "repeat" variables on the Z-machine setting would not always produce a helpful problem message if it also repeated through tables.
Bug fixed (0000314) whereby "repeat" variables were allowed to be changed by "let", which is unsafe. For example,
repeat with x running from 1 to 10:
let x be 1000;
say "[x]."
used to produce the output "1000.", but now produces a problem message instead; the "x" value is protected from changes inside the body of the "repeat".
Bug fixed (0000105b) whereby phrases defined to apply to specific values which are themselves constant phrase names do not work; for example,
To waste time (this is wasting time): do nothing.
To disregard (P - wasting time): do nothing.
caused "disregard wasting time" to fail to work.
Bug fixed (0000229) whereby phrases were forbidden to have two-word names with first word "end", like so:
To end all: end the story saying "It is all over!"
(A hangover to the pre-Pythonesque days when "end ..." phrases were reserved for closing blocks of code; there's no longer any need to reserve these.)
Bug fixed (0000119) whereby descriptions wouldn't work when applied to values stored in the Z-machine as negative numbers between -6 and -1 inclusive. So, for example,
filter to negative numbers of {-2, -1, 0, 1, 2};
would produce a spurious run-time problem instead of {-2, -1}.
Bug fixed (not registered on tracker) whereby definitions like this one:
Definition: a weight (called S) is possible: ...
would result in the temporary value "S" having the kind "object" rather than "weight", making the definition very unlikely to translate properly.
Bug fixed (0000183) whereby use of an undeclared object in a (+ ... +) I6 escape could report a problem at the wrong place, or in some cases crash.
Bug fixed (0000158) whereby references to indexed text values in the past tense, compared against literal ones, would sometimes produce I6 errors. For example,
The message is an indexed text which varies.
Every turn when the message was "send more money": say "Just checking."
Bug fixed (0000219) whereby definitions referring to temporary values in the past tense would result in multiple reports of the same problem message.
Equations, units and arithmetic
Bug fixed (0000328) whereby a value of 0 is printed using the "natural" unit for a given kind even when that natural unit is for use in the singular only. For example,
Population is a kind of value. 1 person (singular) specifies a population.
3 people (plural) specifies a population.
resulted in "0 person", not "0 people".
Bug fixed (0000319) whereby comparisons of truth states with the value "true" would sometimes result in false negatives.
Understanding
Bug fixed (0000208) whereby DROP ... ON ... would be understood as a throwing action rather than a putting action. The Standard Rules now have
Understand "drop [something preferably held] at/against [something]" as
throwing it at.
rather than
Understand "drop [something preferably held] at/against/on/onto [something]"
as throwing it at.
Bug fixed (0000168) whereby problem messages arising from "Understand ... when C" for some invalid condition C would be reported at the wrong sentence, often somewhere in the Standard Rules.
Bug fixed (0000280) whereby multiple Understand lines giving one-word names to things would sometimes allow these words to be combined, and sometimes not, depending on what order they were given in in the source text. For example,
Understand "brass" as the key.
Understand "fob" as the key.
would permit BRASS FOB, BRASS and FOB but not FOB BRASS. This is contrary to a strict reading of the documentation, and (which is more serious) introduces an unnecessary dependency on the order of writing sentences. All four forms of the name now work.
Bug fixed (0000162) whereby, because of names being parsed overly fussily (see 0000280), disambiguation questions would sometimes have the answer refused. An example would be lengthy; see the tracker.
Bug fixed (0000325) whereby spurious problems were generated by attempts to use tokens which gave qualified descriptions of non-object values, e.g.,
A spell is a kind of value. Xyzzy is a spell.
Casting is an action applying to one spell. Understand "cast [possible spell]"
as casting.
Definition: a spell is possible: yes.
Bug fixed (0000292) whereby a test like:
if the player's command matches "[something visible]"...
did not work. (This is a very dubious practice: visibility depends on who is looking, which is not certainly known at this stage in parsing. So it is really best to be more explicit, e.g., "[something which the player can see]".)
Bug fixed (0000192) whereby the parser's response to an odd use of pronouns like this:
Locating is an action applying to one thing.
Understand "locate [a device]" as locating.
Test me with "locate me."
...has been improved.
Bug fixed (0000207) whereby the parser's response to HER, when it can't see the last known female in scope, is improved in cases where the source text also loads the word "HER" with an adjectival meaning, e.g.,
Understand "her" as a thing when the item described is enclosed by a woman.
Bug fixed (0000282) whereby an answer to a disambiguation question which is a fresh request made to somebody else (i.e. ignoring the question) is instead misread as an attempt to disambiguate, with odd consequences, e.g.,
>BOB, GIVE ME THE HAT
Which do you mean, the small red hat or the large yellow hat?
>CAITLIN, GIVE ME THE BOX
There is no reply.
actually not resulting in Caitlin having been asked anything, but from a command BOB, GIVE ME THE CAITLIN, GIVE ME THE BOX HAT - to which, not unreasonably, Bob doesn't know what to say.
Bug fixed (0000191) whereby unnecessary clarification messages would be printed by the parser in certain cases when the choice was made between indistinguishable alternatives. (There's no perfect way to do this, but we found that the new method produced fewer false positives, so to speak, on the corpus of Inform test cases; it usually, but not quite always, reads more naturally than the old way.)
Bug fixed (0000236) whereby the following:
Understand "nothing" as nothing.
produced an internal error instead of a problem message.
Glulx issues, file I/O, figures, sounds
Bug fixed (not on tracker) whereby running Inform on Unix where the "HOME" environment variable is unset, as for instance when running it in a CGI environment, would cause Inform to crash. (It now tries to write files to the root folder, which almost certainly won't work unless it's being run by the root user - madness for CGI usage - and the net effect is that it will fail more gracefully in this case.)
Spacing and printing
Bug fixed (0000305) whereby text with nested, or mismatched, "[" and "]" was allowed, and treated these (usually) as literals; this was bad because it caused syntax-colouring hassles for the user interface applications, made source text more difficult to release to websites, and so on. There are now explanatory problem messages for misuse of "[" and "]".
Bug fixed (0000308) whereby the problem message for an empty text substitution "... [] ..." misleading mentioned use of a comma.
Bug fixed (0000196) whereby lists of objects including closed, empty, light-providing containers did not respect the "Use the serial comma." option.
Bug fixed (0000194) whereby the "[is-are a list...]" and similar text substitutions would wrongly omit "is" when forced to print "nothing" because no objects made the list. This resulted in text like "On the table nothing" instead of "On the table is nothing".
Bug fixed (0000277) whereby the phrase "replace the text X in Y with Z" would always replace using regular-expression matching, rather than literal matching, so that something like this:
replace the text "a" in t with "\(";
would behave unexpectedly because the "" in the replacement text would be read as a literal marker, and so "banana" would become "b(n(n(" not "b(n(n(".
Bug fixed (0000114) whereby the "[one of]...[or]..." text substitution would fail on some cycles if there were large numbers of alternatives present.
Indexing
Bug fixed (0000309) whereby the Actions index omitted "during" clauses.
Bug fixed (0000279) whereby the Actions index omitted commands from the list of "Typed commands leading to this action" when they were verbless commands of the form
Understand "[any room]" as asking for navigation help.
Bug fixed (0000148) whereby, when an optional section is present in the source text, the Contents index displays the "for use with" line, but the Phrasebook index does not, when indexing the contents of the section.
Bug fixed (0000147) whereby the "X understood" variables were unhelpfully spread across the Contents index.
Bug fixed (0000149) whereby relations taking up memory would be incorrectly shown, with garbled names, in the memory usage table at the foot of the Contents index.
Bug fixed (0000255) whereby the Phrasebook index would show the kinds of the terms of relations unhelpfully in the case of some of the built-in ones.
Releasing, bibliographic data, and cBlorb
Bug fixed (not actually registered on bug tracker, though referred to in passing by other reports) whereby the page numbering on released source text websites would be off by one.
Bugs fixed (0000290 and 0000281) whereby projects whose titles include an apostrophe would be difficult to release in various ways. (It's worse if the filename of the Inform project also has an apostrophe in, but that too ought to work in most cases; it seems to be best avoided if releasing along with an interpreter, though. Since there's no requirement for the Inform project to have the same filename as its story title, it may be simplest just to remove any apostrophe used in the filename.)
Bug fixed (0000180) whereby tables in the source text which contained text which contained explicit line and paragraph breaks would be formatted incorrectly in a release of the source text as part of the website.
Bug fixed (0000100) whereby a funny thing happened with headings appearing in the wrong places in a release of the source text as a website, but only on Windows, and... oh, see the Tracker for more.
Bug fixed (0000258) whereby various minor snags with footnotes appeared when releasing the source text as a website.
Bug fixed (0000223) whereby the "Standard" release template resulted in an in-browser interpreter page "play.html" with links assuming that the cover art was in JPEG format, when in fact it might be PNG.
Mac OS X app
-
Problems in running Inform on OS 10.4 (Tiger) and 10.5 (non-Snow Leopard) have been addressed.
-
The Glulx interpreters for the Game pane have been updated to the latest versions: Glulxe 0.4.6 and Git 1.2.8.
-
The "Find" dialog box input focus now works better.
-
Extensions are now created as .i7x files, and an issue with installing extensions has been fixed; similarly with showing errors in extensions.
-
The "Recipe book" menu item (on the Help menu) now directs to the correct page.
-
Fixed an issue displaying the "out of readable memory" page on overflowing Z-machine compilations.
-
Fixed an issue with identifying words when searching.
-
Fixed problem with custom font sets.
Windows app
-
When a project is saved, in order to keep the size of the skein under control, only a limited number of unlocked skein nodes are written out (currently 100). The user is warned of this the first time they open a project with more unlocked skein nodes in it than this. The user should lock any skein threads they want to keep.
-
The "Open Extensions" item in the File menu is now a sub-menu of the available extensions, matching the OS X front end. Extensions built into Inform 7 are shown in smaller text.
-
The old behaviour of the Home and End keys has been restored: they move to the start and end of the line on display, rather than the start and end of the current paragraph.
-
The items on the Format menu now all have keyboard shortcuts.
-
To match the OS X front end, the Transcript pane now has further buttons: "Next in Skein", "Next diff" and "Prev diff". These duplicate functionality from the Replay menu.
-
The Glulx interpreters for the Game pane have been updated to the latest versions: Glulxe 0.4.6 and Git 1.2.8.
Linux app
-
Bug fixed (0000120) which made elastic tabstops work slowly.
-
Updated the interpreters in the Game panel to Glulxe 0.4.6 (plus a patch), Frotz 2.50, and Git 1.2.8.