Some of the material in this paper is ST-80 or ST-V specific, and I have added sections (between horizontal lines) where differences to ST/X or additional facts apply.
Please feel free to send additional classic bugs texts to the original author (Ralph Johnson) or the eXept team, for inclusion in this document.
#add:returns its argument
Array » add:
#become:can ruin your whole day
Every programming system is prone to certain kinds of bugs. A good
programmer learns these bugs, and how to avoid them. Smalltalk is
no different. Smalltalk eliminates lots of bugs that are common in
other languages, such as bugs in linear search algorithms. (Just use
However, it has its own set of classic bugs, which every new Smalltalk
programmer runs into.
There are several reasons to collect classic bugs. The first is that it will help experienced programmers test and debug programs, and can help us design better programs in the first place. Second, if we teach these bugs up front then people should learn to be good programmers faster. Third, perhaps we can redesign the system to eliminate some of these bugs, or we can write checking tools to spot them automatically.
OrderedCollectionare variable-sized classes that grow. They grow by making a copy of themselves and "becoming" the copy. If you add new instance variables to a subclass then you have to make sure that these instance variables get copied, too, or else you will mysteriously lose the values of the instance variables at random points in time.
Smalltalk-80 R4.0 (and probably some earlier versions) has a
copyEmpty: method in
Collection that you are
supposed to override
if you make a subclass of
Collection that adds instance variables.
The solution to this bug is to write a version of
It has been suggested that it would be easy to write a tool that
checked that every new subclass of
Collection that added instance
variables also defined a method for
#become:operation can be quite expensive in direct pointer smalltalk implementations which ST/X is (you don't have to fully understand this :-). Therefore, all collections copy-methods should also copy their container-array, otherwise both the original and the copy might reference the same element-container - leading to much confusion later ...
add:returns its argument, not the receiver, and people usually assume that it returns its receiver. Thus, they write
when they should really write(c add: x) add: y
or elsec add: x; add: y
Note that this is one of the good uses forc add: x. c add: y
yourself, you can write
to make sure that you have the new set.(Set new add: x; add: y; ...; yourself)
Note that there are good reasons why
add: returns its arguments,
and even if there weren't, it is a very, very bad mistake to
add: so that it returns the receiver, and so confuse
every other Smalltalk programmer on the planet.
add: return its argument often keeps you from resorting
to temps, because you can create the argument to
add: on the
fly, and then do other things with it after the
add:. If you
want to access the collection, you can do it with
cascaded messages, as described above.
is a good program, but if you leave out the copy then it isn't.aCollection copy do: [:each | aCollection remove: each]
Mario Wolczko suggested a solution that catches this problem the
instant it occurs (at some performance penalty of course). The
solution is to change the collection classes. Each iteration method
enters that collection into a set of collections being iterated over
(IteratedCollections), executes the block, then removes the collection
from the set. Collections are usually (only?) modified using
basicAt:put:, so these are overriden to check that the collection
is not in IteratedCollections. If it is, an error is signalled. You
can either use this technique all the time, or you can just install
these classes when you are testing and debugging your program. These
changes are packaged in a file
Iterator-check.st that is available on
the Manchester and Illinois servers.
On the Illinois server, it is
The solution is to either provide better documentation, to claim that nobody is allowed to modify copies of collections returned from other objects, or to have objects that don't want their collections modified to return immutable versions of the collections that will give an error if you try to modify them.
where you redefinenew ^ self basicNew initialize
initializein each class to initialize that class's instance variables.
initializeis defined as an instance method:
There are lots of ways to do this wrong. Perhaps the most common is to forget the return, i.e. to write:initialize super initialize "to initialize inherited instance variables" "initialize variables that I define" ...
or:new self basicNew initialize
The result is that you have the class where you want the instance of the class returned fromnew super new initialize
#new. This is a special case of bug #5.
Another error is to forget the superclasses initialization, as in:
This results in a partially initialized object. Be especially careful
when writing a subclass of View; the inherited initialization is
important for proper operation of the view instance.
"initialize variables that I define"
Yet another typical error is to make an infinite loop by writing
If Smalltalk doesn't respond when you think it should, press
CTRL-. (that is "Control" and "." pressed at the same time) to
get the debugger. If the debugger shows a stack of
^ self new initialize
then you know you made this mistake.
Finally, you should be careful to not call "super new initialize", but "basicNew initialize" instead. The reason is that the superclass may itself have redefine the new method to perform initialization. And so, with "super new initialize", we would initialize the object twice, wasting time and perhaps memory.
is perfectly legal Smalltalk, but does dreadful things to your image. This bug could be eliminated if the compiler gave a warning when you assigned to a global variable that contained a class.OrderedCollection := 2
#become:is a very powerful operation. It is easy to destroy your image with it. Its main use is in growing collections (see bug #1), since it can make every reference to the old version of a collection become a reference to the new, larger version. It has slightly different semantics in Smalltalk/V and Smalltalk-80, since
causes every reference to x and y to be interchanged in Smalltalk-80, but does not change any of the references to y in Smalltalk/V.x become: y
Suppose that you want to eliminate all references to an object x.
works fine in Smalltalk/V,
but will cause
every reference to nil to become a reference to x in Smalltalk-80.
This is a sure calamity. You want x to become a new object with no
references, such as in
x become: nil
x become: String new
#become:in ST/X is the same as in Smalltalk-80. However, ST/X provides another method (
#becomeNil) to clear references. Using nil, smallIntegers or living contexts as either receiver or argument to
#become:is not allowed in ST/X.
#become: may be a very time consuming operation in
ST/X (and maybe other smalltalk implementations).
Please avoid the use of it.
If you substitute a new instance of Dictionary or replace, rather than update an association in a pool dictionary, you have to recompile all methods using variables scoped to that Pool.
This is is also annoying when using ENVY, where the methods are under strict control. Perhaps Pool Dictionaries should be be first-class versioned pre-requisites of Classes, just like the class definition.
BTW we are using/VPM 1.4 with ENVY 1.3
1. If you prune & graft a subtree of your class structure you have to make sure that all referencing methods are recompiled. Otherwise you will run (or your customer, because this is only detected at run time) into an Deleted class error message. Thomas Muhr posted a "Bite" a while ago to handle this problem for Smalltalk/V 286.
The same is true, if a classes instance layout changes (for example, when adding an instance variable). Existing instances will remain to exist as instances of the (now obsolete) old class.
You can find removed classes with the MemoryUsageView; look for entries marked as "removed" or "obsolete" (of course, you can also enumerate all behaviors and filter them manually).
in Smalltalk-V andTextPane new open. TextPane new open
in Smalltalk-80, then you will get only one open window, and one forgotten piece of code. This problem has been fixed in Objectworks/Smalltalk R 4.1, so the above code will create two windows as you would expect.aScheduledWindow1 open. aScheduledWindow2 open
The fix for earlier versions of Smalltalk-80 is to use the
openNoTerminate method to open the window, which does not
transfer control to it. A useful trick is to store the new
window in a global variable so you can test it.
Aad Nales says that the fix for Smalltalk/VV is to fork the
creation of the new window.
If this is not what the programmer wants then it is probably necessary
to hack the dispatcher code and remove the dropSenderChain message,
which is the ultimate cause of the problem.
[Textpane open] fork.
openalways returns immediately (and actually forks a new process for the new view, if its a topview), while
openModaldoes not return until the new view is closed.
Originally blocks did not have truly local variables. The block
parameters were really local variables in the enclosing method.
actually had three temporaries, x, y, and z. This leads to bugs
like the following
| x y |
x := 0.
(1 to: 100) do: [:z | x := x + z]
When elements are added to b, the sortBlock is used to
to put them,
but this will change a and b.
| a b |
a := #(4 3 2 1).
b := SortedCollection sortBlock: [:a :b | a someOperation: b].
b addAll: a.
Transcript show: a.
addAll: is OK, but
the a that gets displayed on the transcript will be an integer,
not an array.
Early versions of Smalltalk-80 (2.4 and before?) and of Squeak/Pharo
(pre closure compiler versions)
implemented blocks like this - and Smalltalk/V still does.
Smalltalk/X always implemented real closures right from the beginning
and does not suffer from this problem.
In current Visualworks Smalltalk
implementations, blocks are true closures. You can declare
variables local to a block, and the names of the block parameters are
local to the block. Most people agree that this is a much better
definition of blocks than the original one. Nevertheless, people
planning to use Smalltalk/V should realise that it has a different
semantics for blocks (that comment is from ancient times - today, there are
not many ST/V users left in the world).
This difference can lead to some amusing problems. For example, here
is some code written for ST/V or Squeak by someone who had obviously learned Scheme or Smalltalk/X,
where it works as expected:
The programmer expected that each block would be stored in the array
along with its own value of anIndex.
If anIndex were just a local
variable of the method then this will not work. It assumes that
each execution of the block gets its own version of anIndex, and
Smalltalk/V and old Smalltalk-80 actually make
each execution share
the same version.
| anotherArray aBlockArray |
aBlockArray := Array new: 4.
anotherArray := #(1 2 4 8).
1 to: 4 do: [ :anIndex |
aBlockArray at: anIndex put: [ (anotherArray at: anIndex) * 2 ]].
So, if you are using Smalltalk/V then be careful not to reuse the
names of arguments of blocks unless you know that the blocks are
not going to have their lives overlap. Thus,
is probably OK because
aCollect do: [:i | ...].
bCollect do: [:i | ...].
do: does not store its argument, so the
blocks will be garbage by the time the method is finished.
However, if the first block were stored in a variable somewhere
and evaluated during the execution of the second block then
problems would probably occur.
Note that accepting the method does *not* change the menu. You have to execute the method to change the class variable or class instance variable. Often the initializeMenu method is invoked by the class method initialize. This can lead to the strange effect that you can initialize the menu by deleting the class and filing it in again, but otherwise you don't seem to be able to change the menu (because you haven't figured out that you should really be executing the initializeMenu method).initializeMenu ...
To make matters worse, it is possible that each instance of the controller, or model, or whatever has the menu, stores its own copy of the menu in an instance variable. If that is the case then it is not enough to execute initializeMenu, you also must cause each object to reinitialize its own copy of the menu. It is often easier to just delete the objects and recreate them.
Often a class will have a #flushMenus method to clear out all menus. Typically the method that fetches the menu will check to see if it is nil and invoke initializeMenu if it is. So, flushMenus will just nil out the variable holding the menu. The best way to figure out what is happening is to look at all uses of the variable. Smalltalk experts rarely have problems with this bug, but it often confuses novices.
Caching is a very common technique in Smalltalk for making programs more efficient in both time and space. Caching of menus is one of the simplest uses of caches, and other uses can create more subtle bugs.
add:for Sets and you'll know what I mean: on every
add:the new element is compared to all others resulting into a nonlinear time for adding to Sets.
collect:something from a
SortedCollection, that your result will be sorted as the origin, unless you use the default sortBlock. This is one of the bugs provided by the language vendor
This may make it hard to write a class which shows best performance on
all threee smalltalk implementations.
Assuming that (as a consequence of the Parcplace-Digitalk merge) the old ST/V class library is going to be obsoleted over time, we recommend to ignore the above and use Symbols and Set in a way they are meant to be used (better put preasure on the vendor to fix things instead of writing workarounds ...).
email@example.com (Thomas Muhr) firstname.lastname@example.org (Jan Steinman) email@example.com (Alan Knight) firstname.lastname@example.org (Mario Wolczko) email@example.com (Peter Goodall) Aad Nales <firstname.lastname@example.org> email@example.com (Simon Lewis) firstname.lastname@example.org (Mike Smith) email@example.com (Naci Dai) firstname.lastname@example.org (Dave Robbins) email@example.com (Randy Stafford) Hubert.Baumeister@mpi-sb.mpg.de (Hubert Baumeister) firstname.lastname@example.org (Eliot Miranda) email@example.com (Donald) firstname.lastname@example.org (Amir Bakhtiar) Kurt Piersol <Piersol@Apple.com> email@example.com (Michael Sullivan) firstname.lastname@example.org (Terry) email@example.com (Brent Sterner) firstname.lastname@example.org email@example.com (Nicole Tedesco) firstname.lastname@example.org (Rik Fischer Smoody) email@example.com (Marten Feldtmann) firstname.lastname@example.org (Markus Stumptner)
This bug is especially hard to find, since the method's sourcecode does not
reflect that the array constant has changed - the browser will continue to
show the original sourcecode.
Object subclass:#ArrayLiteralBugDemonstration instanceVariableNames:'' classVariableNames:'' poolDictionaries:'' category:'Examples-Bugs' ! !ArrayLiteralBugDemonstration methodsFor:'bug1'! literalArray ^ #('hello' 'there') ! buggy1 " thats obvious ... " Transcript show:'literalArray initially returns: '; showCR:(self literalArray) displayString. (self literalArray) at:2 put:'wrong'. Transcript show:'literalArray now returns: '; showCR:(self literalArray) displayString. Transcript showCR:'but the source is still:'. Transcript cr. Transcript showCR:(self class compiledMethodAt:#literalArray) source. " ArrayLiteralBugDemonstration new buggy1 " ! buggy2 " less obvious ... hard to track down " Transcript show:'literalArray initially returns: '; showCR:(self literalArray) displayString. Transcript show:'literalArray modified: '; showCR:(self literalArray replaceAll:'hello' by:'bug') displayString. Transcript show:'literalArray now returns: '; showCR:(self literalArray) displayString. Transcript showCR:'but the source is still:'. Transcript cr. Transcript showCR:(self class compiledMethodAt:#literalArray) source. " ArrayLiteralBugDemonstration new buggy2 " ! !
Parserclass. The flag is also found in the Launcher's settings dialog. Set this to true to get immutable array literals, which raise an error when stored into.
Since this is incompatible to Smalltalk-80, the default is false. Also, to the beginner, turning this on may create a bit of confusion if the array protocol is tried in a workspace/inspector, as the object is now presented as an ImmutableArray with a slightly different protocol.
The same mechanism is implemented for literal strings and literal byteArrays.
The binary class compiler stc creates immutable array literals if a corresponding command line option is present.
after you place your comment around the code, as in:... some statement ... "bad statement - disabled by a comment" ... more statements
and accept, you have really reintroduced the previously disabled... " some statement ... "bad statement - disabled by a comment" ... more statements "
Also, comments in the original code may now
be parsed as smalltalk code. Look at:
and the commented version:
foo := foo squared. "some stupid comment"
which now has the comment being compiled as smalltalk code.
foo := foo squared. "some stupid comment"
#comment to the result of sending
to the variable
In most cases, the compiler gives a warning due to unknown identifiers; however, if there happens to be a variable named
the above is perfectly legal and not even a warning is issued.
This is really bad, because you often want to disable debug or print code (which was written during development), but keep that code around for future debugging or as hints for others.
This character sequence was choosen after code studies of existing programs; it was found to not create backward compatibility problems (however, that is just statistics; there might actually be a small number of programs around, for which this is not true)."/ everything after this is a comment
Of course, these comments are not defined in the Smalltalk language standard;
therefore you should be careful in using them,
if you plan to port your code to other Smalltalk dialects.
The ST/X compilers can optionally output a warning, if such a non standard language extension is used.
(BTW: it is well known in the computer society, that comment syntax using the same characters as delimiters is a bad thing. Any upcoming ANSI standard should take care of this).
There is no standard smalltalk solution to this problem.
hereas the receiver of a message send, will result in a method lookup which starts in the class in which the method was defined. Semantically, it is much like a super-send, with the exception, that a super-send starts the search in the superclass of the defining class.
to get the local method - even if the receiver is an instance of a subclass, and the subclass redefinedhere myPrivateMethod.
Keep in mind that here-sends limit the reusability of your classes, since methods using them will never get into any redefined subclass method.Both stc and the incremental compiler offer compilation flags to turn here sends into regular self sends. (of course, you will get a 'using nonstandard features' warning too).
Actually, here sends look much like a function call in other programming languages (except, that if no local target method is found, the search is continued in the superclass).
Notice, that ST/X offers method privacy too. But these will lead to a signal raise at run time instead. Also, method privacy is defined by the target method, while here sends are made up at the caller's side.
add:message, you should try hard to avoid using it. It is very very slow. For this reason, some other systems do not allow
Array » add:at all.
Since arrays are fixed size collections, changing the size implies a
#become: operation. Due to technical reasons, this operation
may be very time consuming in ST/X. If your collection has to
grow, use instances of OrderedCollection instead of arrays. Another alternative is
to use a WriteStream to collect the elements and extract the completed
collection when done (WriteStreams resize their container in big chunks,
avoiding this overhead).
The above is also true for Strings, ByteArrays, FloatArrays and any other fixed size collection.
is up to 500 times slower than both:
a := Array new.
1 to:10 do:[:i |
t := OrderedCollection new.
1 to:10 do:[:i |
a := t asArray
s := WriteStream on:(Array new).
1 to:10 do:[:i |
a := s contents
add:, watch out for other methods which may be
copyWith:method creates a new collection, with one more element. If you use this in a loop, individually adding elements, you may wonder where your time goes.
Methods to be careful about in this respect are (among many others):
In these cases, it is much better to preallocate the target collection, use a WriteStream or work on a temporary resizable container (which is tuned for this type of operation) and convert at the end.
OrderedCollection and others allow you to specify an initial size,
so that resizing is not needed later (but still possible). If you really need an array instance at the end,
convert after building using
asArray as a final message.
s := Set new.
s add:i asFloat
the second piece of code is about 10-20% faster.
(there is some speedup even though Set is already tuned for resizing,
but the effect is not as dramatic as in the example that follows)
s := Set new:10000.
s add:i asFloat
As an extreme example, compare:
a := Array new.
1 to:10000 do:[:i |
a := a copyWith:i
the second is up to 100 times faster.
tmp := OrderedCollection new.
1 to:10000 do:[:i |
a := tmp asArray
See the actual execution times with:
Transcript show:'using copyWith: '.
a := Array new.
1 to:10000 do:[:i |
a := a copyWith:i
Transcript showCR:' ms'.
Transcript show:'using OrderedCollection » add: '.
tmp := OrderedCollection new.
1 to:10000 do:[:i |
a := tmp asArray
Transcript showCR:' ms'.
If you do not know the final size in advance, use a
These are also tuned to handle this kind of growing.
to the (much faster):
s := ''.
s := s , 'more '
the second piece of code is about 5 to 10 times faster.
s := WriteStream on:String new.
s nextPutAll:'more '
When drawing in a view, be careful to keep a reference to the color used.
Otherwise, the garbage collector could free the color, which implies freeing
of the corresponding color index in the displays colormap. The next allocated
color may reuse this index and therefore change the appearance (i.e. the color)
of the previously displayed graph.
is supposed to draw a cross, with different colors used. However,
if a garbage collect occurs after drawing the first line, the first color
may be reclaimed and the same colormap index be reused for the second color.
The first line will therefore be shown in the same color on the screen.
aGC paint:(Color red:15 green:15 blue:100).
aGC displayLineFrom:(0@0) to:(100@100).
aGC paint:(Color red:50 green:50 blue:0).
aGC displayLineFrom:(100@0) to:(0@100).
Avoid this, by keeping references to all used colors somewhere in your view.
Of course, this reference must be kept as long as the view is visible; therefore a method local will usually not be the right place for it. Use an instance variable of an object which lives as long as the view is open.
#become:is a very dangerous operation; it can even crash the system.
MyClassWith2Instvars- with two instance variables, a and b - and another class
MyClassWith1Instvar, which has only one instance variable, called c.
The following method is dangerous, and may lead to a crash:
The problem with the above code is that the compiler translates instance variable accesses
into memory accesses which are indexed by the instance variables number.
I.e. access to a is performed by an access to the first instance variable
(i.e. to the slot with offset 0),
b is accessed at offset 1 and so on.
a := 1. "/ this access is ok
b := 2. "/ still ok
self become:(MyClassWith1Instvar new).
a := 3. "/ mhmh - this now accesses the c instvar
b := 4. "/ CRASH or overwrite of some other object
#become:, the above method continues to access instance variables
self was an instance of
although this is no longer true. Actually, it accesses ``one slot behind'' the changed receiver,
accessing whichever object is found there (this is highly nondeterministic - if by chance, there
happens to be no object, or free memory, nothing happens. If there is some object, it will
Since the very first slot of every object contains a reference to its class, accessing one-behind will very likely lead to the class slot of whichever object happens to be allocated there.
In our concrete example, it is likely that the class slot of some poor innocent victim is modified to 4 which is certainly not a valid class ;-().
Be aware, that the above example is oversimplified:
the instvar-accessing code after the
could be somewhere higher in the calling hierarchy, where it is not known (expected)
that dangerousMethod is ever invoked.
This is one of the very rare cases, where the Smalltalk semantic is not safe - i.e. invalid memory acceses are possible. A possible defence strategy fro the runtime system (VM) could be to abort any currently executed method which is possibly accessing such invalid slots (i.e. all active instance methods of the object). This is currently not done in the VM.
Therefore, this is another reason for banning
#become: from your code.
For above reasons:
Be very very careful in using
#become: - if at all, use it for instances
of the same class only or with instances which have the same layout.
Most code can be easily written without using
#=) method in some class, AND instances of this class are to be placed into a Set or Dictionary (not IdentitySet or IdentityDictionary), you MUST also redefine the
#hashmethod in this class.
The reason is that sets and dictionaries use the hash key to quickly find
an element - and use #= to compare elements further.
If not redefined, the default
#hash method (in
uses the receiver's identityHash-value as a fallback. Thus, if only
is redefined, the objects may compare as equal, but will not return the same
hash keys. Sets and dictionaries will have trouble handling those objects
Therefore, always remember the rule:
Always redefineAs an example, consider a class called person, which redefines equality as:
then, a redefined hash method could be:= somePerson self species == somePerson species ifFalse:[^ false]. self firstName = somePerson firstName ifFalse:[^ false]. self lastName = somePerson lastName ifFalse:[^ false]. ^ true
i.e. the hashValue is best based upon those attributes, which are also taken in the equality test.hash ^ firstName hash + lastName hash
#hashmethod return the same numeric value for instances which compare as equal.
From: email@example.com (William A. Jaeck) Subject: Re: Classic Smalltalk bugs Organization: Ascent Logic Corporation, San Jose, CA Date: Fri, 14 Aug 1992 18:22:27 GMT Here is a Smalltalk bug which bit me just this morning: I had created a subclass of OrderedCollection with an instance method called with:. This is supposed to do the same thing as add:. Then, I implemented a class method called with:with: as with: arg1 with: arg2 ^ self with: arg1; with: arg2 This ended up producing a result as if I had implemented it as ^ self with: arg2 The correct implementation of with:with: is, of course ^ (self with: arg1) with: arg2
From: firstname.lastname@example.org (Rik Fischer SmOOdy) Newsgroups: comp.lang.smalltalk Subject: Classes as a dirty trick to get globals. Ralph: here's one for the list. A common "dirty trick" is to use a class with class methods as an implementation for a "globally known" object which is expected to be unique. This class is never expected to be instantiated: instances would have no capabilities except those inherited by default from #Object. The trick usually works, but it qualifies as design by accident. A tyro might study the object as an example and be misled. (It is important to encourage re-use to keep the visible structure as comprehensible as we can) Future engineers might decide that there needs to be another "instance" of this globally known object. I have seen an example (identity of perpetrators withheld to protect the unconvicted) where the second instance was implmented as a sub-class, with numerous methods over-written to access a different class variable. On Smalltalk V/MAC, a quick scan found (Color MTrap Compiler and LCompiler ) Smalltalk80 V4 comes with DefineOpcodePool but that may have other redeeming social value. The most egregious examples have come with other installed packages. Mr. Manners instead approves of the way Smalltalk80 uses the globally published objects named Transcript, Undeclared, and even Smalltalk itself. Rik Fischer Smoody smOOdynamics Systems Made Outa Objects 2400 NE 25th, Suite 800 Portland, OR 97212 email@example.com 503-249-8300 If the programming environment made it almost automatic to maintain distinguished SortedCollections in the "global" dictionary known as Smalltalk, we would see several examples of SortedCollections installed there, without having elements added to them. Their methodDictionary's would be used as Classes' are now. Naming is important, especially for us humans to get mnemonic value. I personally do not like contrived names that prepend the company of creation, the current author's name, the place in some hierarchy, nor short letters to grandmother. Naming is an art. Practice it well. Let's suppose that they implemented a class CheckBook. On the assumption that you only have ONE CheckBook, they put all the behavior in the class. It had class variables #Balance and #CheckNumber. Later, a second bank account was opened (or they got married?) and a second CheckBook was needed. A subclass CheckBookB was created, with class variables #BalanceB and #CheckNumberB. All methods that accessed either variable were over-written. I think there should have been a class CheckBook with an instance installed as #MyCheckBook (or something). The second instance could then simply be instantiated and installed as #MyCheckBookB. The extra maintainence hassle of installing the first one would have been repaid many times over with the second instantiation.
Newsgroups: comp.lang.smalltalk Date: Thu, 3 Nov 1994 15:57:53 +0100 Sender: SmallTalk programming language discussion <INFO-CLS@DEARN.BITNET> From: Niklas Bjornerstedt <niklas.bjornerstedt@ENTRA.SE> Subject: Common bugs After seeing certain types of bug repeat themselves in many systems I got the idea of compiling a list of the most common ones. The bugs I am interested in are the ones that make it to the end user, not the ones that irritate the developer. To start everything off I'll present my favorites: 1. "self halt" left in the code. 2. delays left in the code. (A friend of mine "optmized" a routine in Object Pascal from 32 to 4 seconds by removing a delay he had forgotten during development.) 3. String handling code unable to handle TwoByteString (ParcPlace Smalltalk). Almost every major package I know of has had bugs in their handling of TwoByteString. 4. Forgetting to test against NULL in arguments and return-values. I could go on but I think this is enough to get things going. Now I invite others to add to this list. I will compile and repost the list if I get any responces. /Niklas ------------------------------------------------------------------------- Niklas Bjornerstedt Entra Data AB, Sweden tel: +46-8-80 97 00 Gustavslundsv. 151 G fax: +46-8-26 04 76 S-161 36 Bromma email: firstname.lastname@example.org
From scrl Mon Jul 20 08:28:42 1992 Relay-Version: version Notes 2.8 87/9/11; site otter.hpl.hp.com From: email@example.com (Simon Lewis) Date: Mon, 20 Jul 1992 07:28:42 GMT Date-Received: Mon, 20 Jul 1992 07:28:42 GMT Subject: Re: Classic Smalltalk bugs Message-ID: <firstname.lastname@example.org> Organization: Hewlett-Packard Laboratories, Bristol, UK. Path: otter!scrl Newsgroups: comp.lang.smalltalk Posting-Version: version Notes 2.8 87/9/11; site otter.hpl.hp.com References: <1992Jul18.email@example.com> Not so much a bug, but a common mistake I think :- (a < b) ifTrue: [a := a + 1]. "Works fine." (a < b) whileTrue: [a := a + 1]. "Is an error. Should be [a < b]." I've seen people do this, and take an awfully long time to figure out why the second one doesn't work, particularly if they *know* the system implements whileTrue:, but have a less than perfect understanding of *how* it works. Simon Lewis, HP Labs, Bristol, UK.