- Pharo-dev - lists.pharo.org

Re: [Pharo-dev] what is exactly flushCache?

by Eliot Miranda

Hi Stephane, the VM caches method lookups in a lookup cache from class,selector pairs to method,primitive pairs, where primitive may be null. This is called the first-level method lookup cache. The JIT VM caches message lookups in machine code, so that a particular piece of machine code exists in a state that invokes a method for a specific class very fast by embedding the class reference in a register load and the target method in a call instruction, and having the target method verify this "cache probe" (this is an "in-line cache). The JIT also caches the translation of a byte coded method to machine code, by hiding a reference to a machine code method in a byte coded method. These caches can be invalidated in several circumstances: 1a. if one adds or removes a method from a class's method dictionary it may change the correct results of a lookup not merely of the class whose dictionary was updated but also subclasses of the class. 1b. if one replaces a method in a method dictionary this changes the target method for a lookup of the selector for the class and subclasses 2. if one wants to rewrite the byte code or literals of a method, for example because a Slot definition has changed, then if the method has been compiled to machine code, the machine code must be discarded before the new code may be executed 1a & 1b are done via Symbol>>flushCache. In response the normal VM flushes its first-level method lookup cache, and the JIT also scans all of machine code looking for inline caches with that selector, and voiding them, reverting each send site for that selector to the "unlinked" state. There used to be confusion in Squeak, which Pharo inherited, that using CompiledMethod>>flushCache was somehow the right way to void caches when updating method dictionaries, flushing the old method in the dictionary, if any, and the new method. It isn't, precisely because adding or removing methods affects the visibility of inherited methods with the same selector. So MethodDictionary code should use Symbol>>flushCache, and only once, on each update of a method dictionary. As a result, the VM will ensure that the necessary send caches are flushed for that selector. 2. is done via CompiledMethod>>flushCache. In response the VM searches the first-level method lookup cache and removes all entries whose target is the method. In addition the JIT discards the machine code for the method, and searches for all send sites with that method's machine code as the target and voids them, reverting them to the unlinked state. The VM must be told to flush the cached state for a compiled method via CompiledMethod>>flushCache and will /try/ and void the state for that method. But it can't always deal with existing activations of that method, because if there are activations running the machine code, that machine code can't merely be thrown away, and can't be replaced because its length may change, depending on literals or byte codes. So this kind of byte coded method manipulation needs to be done with case and some understanding of the total system state. HTH On Tue, Nov 24, 2015 at 9:37 AM, stepharo <stepharo(a)free.fr> wrote: > Clement > > do you know what is flushCache and why we need to invoke it? > > Stef > > -- _,,,^..^,,,_ best, Eliot

Nov. 24, 2015

Re: [Pharo-dev] problem using curl and wget in zero-conf -- explained

by Dale Henrichs

On 11/24/2015 04:30 AM, Ben Coman wrote: > On Tue, Nov 24, 2015 at 8:51 AM, Dale Henrichs > <dale.henrichs(a)gemtalksystems.com> wrote: >> As I mentioned in the previous email, I sometimes have users report that >> they get the following error in their log file when installing >> pharo/GsDevKit-Home: >> >>> /home/travis/build/GsDevKit/GsDevKit_home/shared/pharo/pharo: line 11: >>> /home/travis/build/GsDevKit/GsDevKit_home/shared/pharo >> >> and I assumed that it was some sort of unreported error while downloading >> the pharo-vm (--silent used without --show-error for curl) ... fortunately >> I just started reliably getting this error on travis[1] and was able to >> track down the root cause of the bogus "file does not exist message" to the >> fact that the 32 bit libraries have not been installed[2]. >> >> It's true that the zero-conf script reports info that the 32-bit libraries >> were not installed, but the linux error message in this case is completely >> misleading ... it wasn't until I verified that the file did indeed exist[1] >> that I was able to investigate this further --- in my case I was expecting >> the 32-bit libraries to get installed in a later step and changes to the >> script structure resulted in the 32 bit library install being skipped ... >> >> At the end of the day, I think that error messages should not be suppressed >> by the zero-conf scripts, especially given the fact that the error message >> you get when a download does fail is pretty much the same as you'd get when >> you don't have the 32 bit libraries installed ... >> >> In my case, now that I understand that I cannot rely on a proper error error >> when 32-libraries have not been installed, I will have to see if I can come >> up with a way to provide a better error message for GsDevKit_home users .... >> >> Dale >> >> [1] https://travis-ci.org/GsDevKit/GsDevKit_home/jobs/92839591#L352 >> [2] >> http://askubuntu.com/questions/133389/no-such-file-or-directory-but-the-fil… >> >> >> On 11/23/2015 04:05 PM, Dale Henrichs wrote: >>> >>> >>> On 11/23/2015 03:39 PM, Esteban A. Maringolo wrote: >>>> 2015-11-23 20:25 GMT-03:00 Dale Henrichs >>>> <dale.henrichs(a)gemtalksystems.com>: >>>>> I've been using zero-conf scripts to download the pharo vm for a while >>>>> and >>>>> occasionally the download fails to install the vm in the pharo-vm >>>>> directory >>>>> .... silently ..... with no clue as to what happened. even worse the >>>>> exit >>>>> status is zero so the failure isn't noticed until much later on in my >>>>> install when an attempt is made to actually use the vm .... I do know >>>>> that >>>>> when I see these failures the pharo-vm does get created but is empty >>>>> ... >>>> +1. I've been bitten by this too. >>>> >>>> I would also like the option of seeing the progress of the download of >>>> the sources AND vm, because sometimes I want to know how the download >>>> is going on, speed, etc. >>>> >>>> Regards! >>>> >>>> -- >>>> Esteban >>>> >>> I agree with the progress, but when running as a batch job, you end up >>> with a ton of junk being dumped out into the log file ... getting rid of >>> junk from a log file is how I learned about the problem with --silent:) >>> >>> With regards to the issue I am having --- I have hit a point in my testing >>> (on travis) where I consistently get the: >>> >>>> /home/travis/build/GsDevKit/GsDevKit_home/shared/pharo/pharo: line 11: >>>> /home/travis/build/GsDevKit/GsDevKit_home/shared/pharo/pharo-vm/pharo: No >>>> such file or directory >>> >>> error message yet the following appears to work: >>> >>>> if [ "$OS" == "win" ]; then >>>> PHARO_VM=`find $VM_DIR -name ${VM_BINARY_NAME}.exe` >>>> elif [ "$OS" == "mac" ]; then >>>> PHARO_VM=`find $VM_DIR -name ${VM_BINARY_NAME}` >>>> elif [ "$OS" == "linux" ]; then >>>> PHARO_VM=`find $VM_DIR -name ${VM_BINARY_NAME_LINUX}` >>>> fi >>>> >>>> echo $PHARO_VM >>> >>> since `pharo-vm/pharo` is properly echoed even though the file does not >>> appear to exist? I'm digging deeper .... > I don't have a linux box to check this on, but how about something like... > > if [ `ldd $PHARO_VM 2>/dev/null | grep -c 'not found'` -ne 0 ]; then > echo "Missing library dependencies..." > ldd $PHARO_VM 2>/dev/null | grep 'not found' > echo "Maybe you need to..." > echo " $ sudo dpkg --add-architecture i386" > echo " $ sudo apt-get update" > echo "and for each $MISSING_LIBRARY..." > echo " $ sudo apt-get install $MISSING_LIBRARY:386" > exit 1 > fi > > cheers -ben I tend to agree with the original choice to not fail if the 32-bit libraries are not installed before downloading .... This kind of error exit _should_ be present in the pharo/poharo-ui scripts, so if one attempts to _run_ pharo without having the 32-bit libraries installed, you would get a meaningful error message .... Dale

Nov. 24, 2015

Re: [Pharo-dev] problem using curl and wget in zero-conf

by Dale Henrichs

Actually, I think that you should add the --show-error (with --silent) for curl and --no-verbose for wget, so that you get a more useful error message than "BOOM" ... you also should get a script exit from the curl/wget command as well which is exactly what _should_ happen... Dale On 11/24/2015 02:39 AM, Ben Coman wrote: > On Tue, Nov 24, 2015 at 7:25 AM, Dale Henrichs > <dale.henrichs(a)gemtalksystems.com> wrote: >> I've been using zero-conf scripts to download the pharo vm for a while and >> occasionally the download fails to install the vm in the pharo-vm directory >> .... silently ..... with no clue as to what happened. even worse the exit >> status is zero so the failure isn't noticed until much later on in my >> install when an attempt is made to actually use the vm > So to fail early after the download we should... > if [ ! -f "$VM_DIR/vm.zip" ] ; then > echo "BOOM!" > exit 1; > fi > > (?) cheers -ben > > >> .... I do know that when I see these failures the pharo-vm [directory] >> does get created but is empty ... >> >> I got another failure today and figured that I would dig into this on my end >> to see if I could understand what might be going on .... >> >> The section of the zero-conf script that is suspect is the following: >> >> mkdir $VM_DIR >> $DOWNLOAD_TO$VM_DIR/vm.zip $VM_URL >> >> >> After the $VM_DIR is created (which I know happens), the $DOWNLOAD_TO script >> is run. Recently, I learned that a side effect of using the --silent option >> for curl is that error messages are also suppressed and I see that for curl, >> the `--silent` option is being used: >> >> if [[ `which curl 2> /dev/null` ]]; then >> DOWNLOAD="curl --silent --location --compressed "; >> DOWNLOAD_TO="$DOWNLOAD --output "; >> elif [[ `which wget 2> /dev/null` ]]; then >> DOWNLOAD_TO="wget --quiet --output-document="; >> DOWNLOAD="$DOWNLOAD_TO-"; >> else >> echo "Please install curl or wget on your machine"; >> exit 1 >> fi >> >> >> For curl if you want to turn off the progress meter (a good thing) and >> continue to report errors, you need to use --show-error in conjunction with >> --silent (as described in the man page) >> >> I think the same thing applies to wget, --quiet turns off error messages and >> progress .... you probably should be using --no-verbose to continue to get >> error messages ... >> >> Dale >> >>

Nov. 24, 2015

Re: [Pharo-dev] News about FastTable

by Thierry Goubier

Hi Cyril, Le 24/11/2015 11:58, Ferlicot D. Cyril a Ã©crit : > Le 18/11/2015 10:53, Ferlicot D. Cyril a Ã©crit : >> Hi everyone! >> >> I just wanted to say that I added a first version of Tree in FastTable >> and I corrected some bugs. >> >> If someone want to try it and give some feedback you can see some >> examples on the latest Pharo 5 in FTExamples class. >> >> With Usman we also worked on the glamour version of FastTable begun by >> Esteban. >> You can see some examples of FastList, FastTable, FastOutlineList and >> FastTree in the GLMExamplesBrowser after executing this: >> >> (ConfigurationOfFastTable project version: #stable) load: #glamour >> >> Every feedback is welcome :) >> > > Hi, > > I took into account your feedback. > > I corrected the broken examples. > > I copied the Glamour package into glamour repository. I did not produced > a new configuration yet. > > Now the FastTree selection is updated when we collapse/extend an item. > The selection of collapsed item is not saved because with the current > implementation of FastTable that would be really complex. I think that > if someone want to have a widget with this feature that should be a > subclass of FTTreeDataSource and not a basic feature. Agreed. > And I did some small changes. > Selection probably doesn't work very well: I got a DNU in updateSelectionWithCollectBlock: (running exampleTree1, and trying the select in subtree then collapse) In Morphic-Widgets-FastTable-CyrilFerlicot.133 (changing collectAsSet: by collect: gives the correct behavior on the selection). Thierry

Nov. 24, 2015

Re: [Pharo-dev] what is exactly flushCache?

by Marcus Denker

> On 24 Nov 2015, at 14:37, stepharo <stepharo(a)free.fr> wrote: > > flushCache The comment says it: "Tell the interpreter to remove all references to this method from its method lookup cache, if it has one. This primitive must be called whenever a method is redefined or removed.â Marcus

Nov. 24, 2015

About TBehavior

by stepharo

Hi I was reading the code of TBehavior and I'm puzzled. I do not get why - initialize does not call superclass: aClass methodDictionary: mDict format: fmt - what is this comment about not sending messages.... and doing it just after :( TBehavior >> initialize "moved here from the class side's #new" super initialize. self superclass: Object. "no longer sending any messages, some of them crash the VM" self methodDict: self emptyMethodDictionary. self setFormat: Object format. self traitComposition: nil. self users: IdentitySet new. TBehavior >> superclass: aClass methodDictionary: mDict format: fmt "Basic initialization of the receiver. Must only be sent to a new instance; else we would need Object flushCache." self superclass: aClass. self setFormat: fmt. self methodDict: mDict. self traitComposition: nil I really think that I would like to revisit the traits introduction in the kernel. Stef

Nov. 24, 2015

what is exactly flushCache?

by stepharo

Clement do you know what is flushCache and why we need to invoke it? Stef

Nov. 24, 2015

Re: [Pharo-dev] Rules

by Yuriy Tymchuk

> On 24 Nov 2015, at 17:29, Ben Coman <btc(a)openinworld.com> wrote: > > On Tue, Nov 24, 2015 at 11:45 PM, Yuriy Tymchuk <yuriy.tymchuk(a)me.com <mailto:yuriy.tymchuk@me.com>> wrote: >> >> On 24 Nov 2015, at 14:10, Ben Coman <btc(a)openinworld.com> wrote: >> >> On Tue, Nov 24, 2015 at 4:32 PM, Nicolas Anquetil >> <nicolas.anquetil(a)inria.fr> wrote: >> On 23/11/2015 22:14, stepharo wrote: >> Le 22/11/15 22:11, Yuriy Tymchuk a Ã©crit : >> >> >> Hi. >> >> The rule is correct. Sending a different message to super may be >> confusing. In this case you are in âprintOn:â method and you are sending >> âprintOn:nextPutAll:â to super. The idea is that you have to send >> âprintOn:nextPutAll:â to self and then either it will invoke the super one, >> you there will be an overridden method in your current class. >> >> exactly but do you expect a newbie to understand anything with the current >> explanation? >> I do not. >> >> Rewrite super messages to self messages when both refer to same method" >> >> >> What about : >> >> >> Sorry :) both of these make my head hurt. I think Stef's was clearer. >> The useful part there is that it reports the actual identifier causing >> the problem, which makes it more obvious this is not what was meant, >> from which even a newbie may infer the missing period. >> >> "'super' should call the same selector. Send different selector to 'self' >> then use 'super' in this one." >> ? >> >> >> btw, is this just commonly how super is used, or is it really bad to >> call a different super method? >> I'm sure I've done that on occasion. >> >> >> It makes harder to extend your code. Because if method `a` calls method `b`, >> then even if method `b` is not defined in your class, I can subclass and >> redefine method b. Otherwise I have to subclass and redefine `a` although >> `a` is fine, but I want to run my `b` and not the one of superclass. >> >> >> Uko >> >> > > Thanks Uko. I was considering a case where my new subclass has > redefined 'b', but I want my 'a' to avoid my 'b' and use the > superclass one. > > (Sorry I can't remember the use case) > cheers -ben Hi, no rule is perfect, and there are definitely cases that work the other way around. But we try to focus on the best practices. Also I think this influences the maintainability of code. Because in the suggested by the rule, you have method `a` with one responsibility and it uses method `b` with another responsibility. Now when you call method `b` from the super class, all logic becomes more complicated. Uko > > >> cheers -ben >> >> >> or: >> >> "should not send a different message to super (send different message to >> self and then call super from it)" >> ? >> >> >> >> nicolas >> >> >> Sucks >> >> >> Secondly, there is an issue with âsends unkown selectorâ. Iâve also >> noticed it recently. I think that it happens because the rule has a special >> cace, and I messed up something with refresh. This is the next thing on my >> TODO list. >> >> Cheers. >> Uko >> >> >> On 22 Nov 2015, at 16:33, Marcus Denker <marcus.denker(a)inria.fr> wrote: >> >> >> On 22 Nov 2015, at 12:20, stepharo <stepharo(a)free.fr> wrote: >> >> With the following (broken) definition >> >> >> : aStream >> super printOn: aStream >> aStream nextPutAll: '(', faces printString , ')' >> >> I got this strange rule: >> >> Rewrite super messages to self messages when both refer to same >> method >> >> I do not get it and it looks really suspicious to me. >> Any suggestions? >> >> >> The rule is meant to catch the case when when âsuperâ can be âselfâ as >> there is no #printOn: >> defined in the class. >> >> But here it looks like it is wrong: it would be an infinite loop. >> >> I think it could be related to this other bug: I have seen that the >> rules are not correctly updated >> when you add methods. (I wanted to add an issue for that but did not >> yet). E.g. I often get >> a âsends unkown selectorâ for self sends and then I check â> method is >> there, I just added it! >> >> So here this means that you added a new printOn:, the rules do not >> update, and it thinks >> that there is no printOn:. >> >> So the real bug is that added methods need to be taken into account by >> QA. >> >> Marcus >> >> >> >> >> >> >> >> -- >> Nicolas Anquetil >> RMod team -- Inria Lille

Nov. 24, 2015

Re: [Pharo-dev] Improving API of Date, Month....

by stepharo

Hi guys I made the mistake to think that it was a good move to improve and make the core more full (and complex in the past). Now what I would love is the following. - keep and reduce if necessary the core date classes (only used internally by the system) - have nice packages that we can load to represent time / calendar -- chronos -- aconcagua -- or a new one That do it the right way: with locale and so on. Why? because it can be complex and verbose and we want to have a small core (for many different reasons). About the durationFormatter I think that we should definitively have more strategies to represent dates and other within a nice date/calendar package. I would love that someone propose something to get the nice extensible Calendar/Date package Stef Le 24/11/15 16:25, Esteban A. Maringolo a Ã©crit : > 2015-11-24 12:14 GMT-03:00 Skip Lentz <skip.lentz(a)inria.fr>: >> Hi, >> >> I also had an idea for a method which returns the most significant unit of a Duration. >> >> For example, you have a duration of 432 days, 3 hours, 21 minutes, 5 seconds, etc., and >> it would return you "432 days" (or just â1 yearâ). I encountered this when wanting to create >> a time indication on e.g. a commit or a comment. Itâs much more readable and user-friendly >> than a timestamp. >> >> Would this be a nice addition to the Duration API? > It would be nice to have. But why not something like a > DurationFormatter? that in turn collaborates with the Locale to ge the > words for "Year, Month, Week" in the current language. > > Regards! > >

Nov. 24, 2015

Re: [Pharo-dev] Kind of RBRenameMethodRefactoring but for closures?

by Mariano Martinez Peck

On Tue, Nov 24, 2015 at 3:04 AM, Thierry Goubier <thierry.goubier(a)gmail.com> wrote: > Hi Mariano, > > Le 22/11/2015 23:31, Mariano Martinez Peck a Ã©crit : > >> >> >> ... > >> >> You know what is funny? I run all AST-Tests-Core from Pharo and NONE >> halt on #createBlockFor: ... >> I even run all those from Refactoring-Tests-Core. >> *So..it seems there is not tests that covers the usage of >> #createBlockFor: :(* >> > > Ok. I'll have a scan of all the RB tests and add additional tests then. > There is an issue John pointed to Nicolai and me that needs solving as well. > > OK. Great. Once you do it let me know and I will see if I can port and make them work for GemStone. > BTW, do you image other cases like this? >> >> >> You mean storing those RB rules? I try to collect them if I see some >> (or participate to), or remember where I see them. >> >> >> No, I mean the compilation of block closures that would try to bind >> "self". Anyway, I will search if I find more. >> > > Thanks. > > I used to know where the RBRewriteTree chapter was in the Pharo >> books, but it seems to have disappeared, so I have recreated a copy >> that I'll try to update. >> >> >> >> Thanks, please share it with me when you made it because I could not >> find it weeks ago. >> > > My fork of the Pharo for the enterprise book has it: > > https://github.com/ThierryGoubier/PharoForTheEnterprise-english > > But I don't have a CI building it yet. > > Thanks! I will read it today! -- Mariano http://marianopeck.wordpress.com

Nov. 24, 2015