- Pharo-dev - lists.pharo.org

Re: [Pharo-project] [squeak-dev] Re: Alien & Squeak FFI issues on Snow Leopard.
by John M McIntosh Sept. 30, 2009

Sept. 30, 2009

The reason for the SqueakPluginsBuiltInOrLocalOnly was to prevent the VM from looking in about 100 places for a plugin before deciding to use the one built into the VM this saved several 10 of ms in launch time. At the time there was also a transition period between PowerPC and Universal plugins, the historical algorithm would try to load powerpc plugins lying loose in a developer's various libraries, versus using the newly complied ones supplied by the VM. So setting SqueakPluginsBuiltInOrLocalOnly to NO might lead to other interesting issues, versus for example doing a 'ln -s /usr/lib/ libc.dylib libc.dylib in the VM's resource folder. I will look at changing sqMacUnixExternalPrims.c to allow absolute paths, and also consider the changes Ian made lately to reduce the amount of searching that the original UnixExternalPrims.c did. On 2009-09-30, at 9:04 AM, Eliot Miranda wrote: > Craig Latta found a simple work-around which is to set the VM's > Info.plist variable "SqueakPluginsBuiltInOrLocalOnly" to false, so > that libraries are found on Snow Leopard. Up until now things have > been being found through vagaries if the Mac OS X dlopen call which > would, I think, search the paths in an executable used to find > dlls. i.e. prior to Snow Leopard if you tried to run something in > libSystem.B.dylib and you traced the logic in > sqMacUnixExternalPrims.c you'd find that dlopen was "finding" > libSystem.B.dylib in VM.app/Contents/Resources, even though it is > not there. > > Personally I think that sqMacUnixExternalPrims.c is ripe for a > rewrite. It is torturously complex. But Craig's workaround works > for us and takes the pressure off. > > P.S. I'm not on the Pharo list. Can someone forward this for me? -- = = = ======================================================================== John M. McIntosh <johnmci(a)smalltalkconsulting.com> Twitter: squeaker68882 Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com = = = ========================================================================

1 0

[Pharo-project] Alien and special objects Array
by Torsten Bergmann Sept. 30, 2009

Sept. 30, 2009

Only the lates core counts ;) In Core #10463 we have: BlockContext>>fixTemps -> the old implementation BlockClosure>>fixTemps -> the empty no-op for compatibility Bye T. -- Neu: GMX Doppel-FLAT mit Internet-Flatrate + Telefon-Flatrate fÃ¼r nur 19,99 Euro/mtl.!* http://portal.gmx.net/de/go/dsl02

2 1

[Pharo-project] [update] #10464
by Marcus Denker Sept. 30, 2009

Sept. 30, 2009

10464 ------ Issue 1255: isMorphic left overs in Debugger and ChangesOrganizer Issue 1061: Not all FreeType fonts load properly in Windows Issue 1173: DisplayText have no accessor #form: Issue 1131: 2 Bad comments -- Marcus Denker - http://marcusdenker.de PLEIAD Lab - Computer Science Department (DCC) - University of Chile

1 0

[Pharo-project] Assignment operator in new core packages
by Torsten Bergmann Sept. 30, 2009

Sept. 30, 2009

Classes like "SHA1" (now included in core) use _ again instead of := Shouldnt we clean up such packages? Thx T. -- GRATIS fÃ¼r alle GMX-Mitglieder: Die maxdome Movie-FLAT! Jetzt freischalten unter http://portal.gmx.net/de/go/maxdome01

4 3

[Pharo-project] Quickly browse through large repositories
by Lukas Renggli Sept. 30, 2009

Sept. 30, 2009

When clicking on a package in a large repository the browser usually takes ages to refresh. The following fix in the inbox fixes this problem. Lukas Name: MonticelloGUI-lr.5 Author: lr Time: 30 September 2009, 9:04:31 am UUID: fd2ef6fb-d926-49a6-a5d6-557e4218576b Ancestors: MonticelloGUI-stephane_ducasse.4 - quickly browse through large repositories -- Lukas Renggli http://www.lukas-renggli.ch

2 1

[Pharo-project] Alien & Squeak FFI issues on Snow Leopard.
by John M McIntosh Sept. 30, 2009

Sept. 30, 2009

Someone finally managed to fire up a Sophie app http://www.opensophie.org/ on Snow Leopard and immediately found an issue that also affects Squeak users who do fascinating things with FFI The innocent FFI call was apifstat: afileNo statBuffer: buffer <cdecl: long 'fstat' (long MacStat* ) module: 'libc.dylib'> ^self externalCallFailed Now it seems BEFORE Snow Leopard *somehow* the Squeak external primitive call api which is part of the VM to assist in loading plugins or finding function entry points in executable binaries or libraries would return the entry point for fstat by *somehow* finding libc.dylib which is via /usr/lib/libc.dylib -> libSystem.dylib -> libSystem.B.dylib Interestingly this is found implicitly since the mac carbon vm lookup logic never looks in /usr/lib/ Now in Snow Leopard it seems they do less implicit searching, perhaps a security concern? Then it fails. In the past the Unix VM would search a hundred(sic) more places, but I understand less today, but it might still work? Maybe /usr/lib is in the hit list? This issue also applies to Alien because it calls back to the VM for it to find the module and function entry point. Since I have received a request to ensure that doing an explicit full path should be allowed in the squeak carbon macintosh vm. Then a change would be: apifstat: afileNo statBuffer: buffer <cdecl: long 'fstat' (long MacStat* ) module: '/usr/lib/libc.dylib'> ^self externalCallFailed I'll look at doing this shortly. However the workaround is to go to Sophie.app/Contents/Resources and do a ln -s /usr/lib/libc.dylib libc.dylib to create a symbolic link in the Resources directory for libc.dylib since we do explicitly look for libc.dylib in the Resources directory. Obviously if you have a FFI call like above where you are calling a system library Oracle? for example. You might find it no longer works with Snow Leopard on the Macintosh, so you should check. -- = = = ======================================================================== John M. McIntosh <johnmci(a)smalltalkconsulting.com> Twitter: squeaker68882 Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com = = = ========================================================================

2 1

[Pharo-project] [update] closure fixes :)
by StÃ©phane Ducasse Sept. 30, 2009

Sept. 30, 2009

Thanks thanks thanks eliot and mike. Now I will check if we did not rollbacked some fixes I did and some pending changes.... And we should soon to ReleaseCandidate. Stef

3 3

Re: [Pharo-project] [squeak-dev] Mac carbon VM goes to 4.2.2Beta1U
by John M McIntosh Sept. 29, 2009

Sept. 29, 2009

Ok, I've altered the master Squeak 4.2.2beta1U.app.zip and added the signature file. Also I've attached it. People who have made copies (like for the pharo site) might want to re- download it to ensure you've the correct zip file. found via http://smalltalkconsulting.com/squeak.html On 2009-09-29, at 9:12 AM, Bert Freudenberg wrote: > > On 27.09.2009, at 23:49, John M McIntosh wrote: >> >> I've also uploaded the new VM as Squeak 4.2.2beta1U.app.zip >> found via >> http://smalltalkconsulting.com/squeak.html > > Is it intentional there is no "Squeak VM Opt.sig" anymore? > > - Bert - > > -- = = = ======================================================================== John M. McIntosh <johnmci(a)smalltalkconsulting.com> Twitter: squeaker68882 Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com = = = ========================================================================

1 0

[Pharo-project] Update 10458 fails with MNU
by Martin McClure Sept. 29, 2009

Sept. 29, 2009

Downloaded 09.09.3, install fresh, launch, System->update. Runs for a while, then fails while loading 10458. Failure is MNU. In [] in ScriptLoader>>loadTogether:merge: the code [:fn | loader addVersion: (self loadVersionFromFileNamed: fn)] Is failing, #loadVersionFromFileNamed: is answering nil when fn is 'Gofer-lr.48.mcz'. And, indeed my package-cache does not contain 'Gofer-lr.48.mcz', but it does have Gofer-lr.51.mcz. Regards, -Martin

4 5

[Pharo-project] Morph extension
by Gary Chambers Sept. 29, 2009

Sept. 29, 2009

After some experimenting I can get over 10% perfromance improvement with window operations when fillStyle; borderStyle; hResizing; vResizing are made as instance variables of Morph, rather than going through extension. I thought Marcus has mentioned this some time ago, but no progress? There is obviously an issue with name collision/shadowing but not many methods in core need changing to accomodate. Whilst on the subject I am wondering on everyone's opinion on merging the behaviour of color/borderColor/borderWidth with that of fillStyle... sure to be some gains there! Regards, Gary

5 12