- Pharo-dev - lists.pharo.org

[Pharo-project] UUIDs not so unique

by Igor Stasenko

On 20 July 2011 18:58, Eliot Miranda <eliot.miranda(a)gmail.com> wrote: > Hi Igor, (Hi Phillipe) > Â Â Â yes, I see the bug in the linux Cog VMs. Â I went to > http://code.google.com/p/pharo/issues/detail?id=4541Â but I don't see any > fix. Â Can you point me to the fix and I'll integrate and roll out new VMs > ASAP. > TIA > Here: https://gitorious.org/cogvm/blessed/commit/4a3f8bf9f560c19b167ee1b3d5ab6c52… > On Wed, Jul 20, 2011 at 8:51 AM, Igor Stasenko <siguctua(a)gmail.com> wrote: >> >> On 20 July 2011 16:27, Torsten Bergmann <astares(a)gmx.de> wrote: >> > Igor wrote: >> >>Feel free to submit an additional test for UUID to prevent this >> >>regression in a future. >> > >> > I already provided such a test in my previous post: >> > >> > >> > http://lists.gforge.inria.fr/pipermail/pharo-project/2011-July/051481.html >> > >> > and now there is a fixed issue ready for integration >> > >> > http://code.google.com/p/pharo/issues/detail?id=4541 >> > >> Thanks, Torsten. >> >> > Nothing more to do on Pharo side than integration this test case >> > into one of the next updates. >> > >> > >> > But the VM maintainers should fix the issue for the VM >> > Phillipe used. >> > >> i will check the unix configs for Cog VMs. >> But since Philippe using Eliot's builds, then he should wait till >> Eliot integrate it (hoping that Eliot reads this mailing list and >> corresponding topic). >> >> > Bye >> > T. >> > -- >> > NEU: FreePhone - 0ct/min Handyspartarif mit Geld-zurÃ¼ck-Garantie! >> > Jetzt informieren: http://www.gmx.net/de/go/freephone >> > >> > >> >> >> >> -- >> Best regards, >> Igor Stasenko AKA sig. >> > > > > -- > best, > Eliot > -- Best regards, Igor Stasenko AKA sig.

July 20, 2011

Re: [Pharo-project] [update 1.4] #14050

by StÃ©phane Ducasse

July 20, 2011

Re: [Pharo-project] UUIDs not so unique

by Eliot Miranda

Hi Igor, (Hi Phillipe) yes, I see the bug in the linux Cog VMs. I went to http://code.google.com/p/pharo/issues/detail?id=4541 but I don't see any fix. Can you point me to the fix and I'll integrate and roll out new VMs ASAP. TIA On Wed, Jul 20, 2011 at 8:51 AM, Igor Stasenko <siguctua(a)gmail.com> wrote: > On 20 July 2011 16:27, Torsten Bergmann <astares(a)gmx.de> wrote: > > Igor wrote: > >>Feel free to submit an additional test for UUID to prevent this > >>regression in a future. > > > > I already provided such a test in my previous post: > > > > > http://lists.gforge.inria.fr/pipermail/pharo-project/2011-July/051481.html > > > > and now there is a fixed issue ready for integration > > > > http://code.google.com/p/pharo/issues/detail?id=4541 > > > Thanks, Torsten. > > > Nothing more to do on Pharo side than integration this test case > > into one of the next updates. > > > > > > But the VM maintainers should fix the issue for the VM > > Phillipe used. > > > i will check the unix configs for Cog VMs. > But since Philippe using Eliot's builds, then he should wait till > Eliot integrate it (hoping that Eliot reads this mailing list and > corresponding topic). > > > Bye > > T. > > -- > > NEU: FreePhone - 0ct/min Handyspartarif mit Geld-zurÃ¼ck-Garantie! > > Jetzt informieren: http://www.gmx.net/de/go/freephone > > > > > > > > -- > Best regards, > Igor Stasenko AKA sig. > > -- best, Eliot

July 20, 2011

Re: [Pharo-project] why isEmptyOrNil is not defined in Object?

by StÃ©phane Ducasse

July 20, 2011

Re: [Pharo-project] UUIDs not so unique

by Igor Stasenko

On 20 July 2011 17:31, Levente Uzonyi <leves(a)elte.hu> wrote: > On Wed, 20 Jul 2011, Igor Stasenko wrote: > >> I changed the MakeUUID in sqUnixUUID.c to look like following: > > This change is useless IMHO. Instead of stopping the compilation, either > return 0 from the function, so the primitive will fail, or fix the included > header files for Darwin, which has uuid_generate > (http://developer.apple.com/library/mac/documentation/Darwin/Reference/ManPa… > ). But the best solution is using the Mac OS platform files instead of the > Unix ones for Mac OS. > Right, that's why i pointed Esteban to use mac-specific functions. Well, the problem was initially reported for unix builds. And why stopping compilation useless? It indicating the configuration error, while if we do as you proposing, then prim will always fail and UUID will work.. but then its a question, why we need uuid plugin at all, if it's not using any function for generating uuids? I prefer to have compilation error, than useless and malfunctioning plugin which does nothing. > > Levente > >> >> >> >> int MakeUUID(char *location) >> { >> Â uuid_t uuid; >> >> #if defined(HAVE_UUIDGEN) >> Â uuidgen(&uuid, 1); >> #elif defined(HAVE_UUID_GENERATE) >> Â uuid_generate(uuid); >> #else >> #error "No uuid function provided" >> #endif >> >> Â memcpy((void *)location, (void *)&uuid, sizeof(uuid)); >> Â return 1; >> } >> >> >> and during compiling cocoa vms it fails with error. >> And before change the code was looking like following: >> >> int MakeUUID(char *location) >> { >> Â uuid_t uuid; >> >> #if defined(HAVE_UUIDGEN) >> Â uuidgen(&uuid, 1); >> #elif defined(HAVE_UUID_GENERATE) >> Â uuid_generate(uuid); >> #endif >> >> Â memcpy((void *)location, (void *)&uuid, sizeof(uuid)); >> Â return 1; >> } >> >> which means that if none HAVE_UUIDGEN or HAVE_UUID_GENERATE is >> defined, then there is nothing got called >> and memcpy then returns a random values on stack, copied from uuid to >> locaiton. >> >> And first version of that file just contains : >> >> >> int MakeUUID(char *location) >> { >> Â uuid_t uuid; >> Â uuid_generate(uuid); >> Â memcpy((void *)location, (void *)&uuid, sizeof(uuid)); >> Â return 1; >> } >> >> and then changed to: >> >> #if defined(__NetBSD__) >> Â uuidgen(&uuid, 1); >> #else >> Â uuid_generate(uuid); >> #endif >> >> and then changed to: >> >> #if defined(HAVE_UUIDGEN) >> Â uuidgen(&uuid, 1); >> #elif defined(HAVE_UUID_GENERATE) >> Â uuid_generate(uuid); >> #endif >> >> Blame: >> Merge Levente's linux UUIDPlugin fix. Fix tickerSleepCycle decl in >> heartbeat. Restore -O2 optimization for the interpreter. >> >> Author: eliot >> Date: Â 5 months ago >> Commit: 757297f62c32eaf2af7f26ce66aa6499df776749 >> >> >> >> -- >> Best regards, >> Igor Stasenko AKA sig. >> >> > > -- Best regards, Igor Stasenko AKA sig.

July 20, 2011

Re: [Pharo-project] UUIDs not so unique

by Igor Stasenko

On 20 July 2011 16:27, Torsten Bergmann <astares(a)gmx.de> wrote: > Igor wrote: >>Feel free to submit an additional test for UUID to prevent this >>regression in a future. > > I already provided such a test in my previous post: > > http://lists.gforge.inria.fr/pipermail/pharo-project/2011-July/051481.html > > and now there is a fixed issue ready for integration > > http://code.google.com/p/pharo/issues/detail?id=4541 > Thanks, Torsten. > Nothing more to do on Pharo side than integration this test case > into one of the next updates. > > > But the VM maintainers should fix the issue for the VM > Phillipe used. > i will check the unix configs for Cog VMs. But since Philippe using Eliot's builds, then he should wait till Eliot integrate it (hoping that Eliot reads this mailing list and corresponding topic). > Bye > T. > -- > NEU: FreePhone - 0ct/min Handyspartarif mit Geld-zurÃ¼ck-Garantie! > Jetzt informieren: http://www.gmx.net/de/go/freephone > > -- Best regards, Igor Stasenko AKA sig.

July 20, 2011

Re: [Pharo-project] UUIDs not so unique

by Philippe Marschall

On 07/20/2011 05:08 PM, Casimiro de Almeida Barreto wrote: > Em 20-07-2011 11:29, Philippe Marschall escreveu: >> On 07/20/2011 03:59 PM, Torsten Bergmann wrote: >>> (...) >> Did the bug eat your code? > Are you a professional developer? If so you must have got stuck in bugs > before. MS bugs, Linux bugs, Mac OS bugs, compiler bugs, jdk bugs, > library bugs... > > I have been stuck on bugs. And I can tell you for sure that pharo/squeak > communities are extremely responsive. You point a bug and solution comes > in hand. I can not tell the same about "commercial software > manufacturers" (just try "online support" from the major OS supplier in > the market... or file a bug and track how long it takes to be fixed). > And remember, contrary to pharo, you have to purchase "visual this or > that" in order to use such "marvelous development suite". Even in open > source communities: you have a trouble with Linux kernel you fill a > report in bugzilla and it takes sometimes years to be closed. So, I > can't understand: > > a) your over reaction It ate my code. > b) your insistence in a subject that has proved itself exhausted. > > By the way, the bug didn't "eat your code". Just put the primitive > expression inside double quotes and your code will work all right > (supposing it's correct). Yeah right. I already have an SCM repository (Seaside 3) with several versions with the same UUID. First I have to write a tool that asses the damage which means pulling all my open source code including the full Seaside 3 repository from SqueakSource and search for duplicate UUIDs. Then I have to hope and pray that the corruption doesn't go too far back. Then for all the broken packages I have to load the last correct code. Then for each of them I have to load the latest code without Monticello and commit it with a different version number. Then I have to remove all broken the versions with screws everybody who already used them as ancestors. Cheers Philippe

July 20, 2011

Re: [Pharo-project] UUIDs not so unique

by Levente Uzonyi

On Wed, 20 Jul 2011, Igor Stasenko wrote: > I changed the MakeUUID in sqUnixUUID.c to look like following: This change is useless IMHO. Instead of stopping the compilation, either return 0 from the function, so the primitive will fail, or fix the included header files for Darwin, which has uuid_generate (http://developer.apple.com/library/mac/documentation/Darwin/Reference/ManPa… ). But the best solution is using the Mac OS platform files instead of the Unix ones for Mac OS. Levente > > > > int MakeUUID(char *location) > { > uuid_t uuid; > > #if defined(HAVE_UUIDGEN) > uuidgen(&uuid, 1); > #elif defined(HAVE_UUID_GENERATE) > uuid_generate(uuid); > #else > #error "No uuid function provided" > #endif > > memcpy((void *)location, (void *)&uuid, sizeof(uuid)); > return 1; > } > > > and during compiling cocoa vms it fails with error. > And before change the code was looking like following: > > int MakeUUID(char *location) > { > uuid_t uuid; > > #if defined(HAVE_UUIDGEN) > uuidgen(&uuid, 1); > #elif defined(HAVE_UUID_GENERATE) > uuid_generate(uuid); > #endif > > memcpy((void *)location, (void *)&uuid, sizeof(uuid)); > return 1; > } > > which means that if none HAVE_UUIDGEN or HAVE_UUID_GENERATE is > defined, then there is nothing got called > and memcpy then returns a random values on stack, copied from uuid to locaiton. > > And first version of that file just contains : > > > int MakeUUID(char *location) > { > uuid_t uuid; > uuid_generate(uuid); > memcpy((void *)location, (void *)&uuid, sizeof(uuid)); > return 1; > } > > and then changed to: > > #if defined(__NetBSD__) > uuidgen(&uuid, 1); > #else > uuid_generate(uuid); > #endif > > and then changed to: > > #if defined(HAVE_UUIDGEN) > uuidgen(&uuid, 1); > #elif defined(HAVE_UUID_GENERATE) > uuid_generate(uuid); > #endif > > Blame: > Merge Levente's linux UUIDPlugin fix. Fix tickerSleepCycle decl in > heartbeat. Restore -O2 optimization for the interpreter. > > Author: eliot > Date: 5 months ago > Commit: 757297f62c32eaf2af7f26ce66aa6499df776749 > > > > -- > Best regards, > Igor Stasenko AKA sig. > >

July 20, 2011

Re: [Pharo-project] UUIDs not so unique

by Igor Stasenko

On 20 July 2011 17:15, Esteban Lorenzano <estebanlm(a)gmail.com> wrote: > he, we made same change... you are going to see a conflict in blessed sources :) > No. i didn't committed it. I just put an obvious missing code path to see what's wrong there. So your's will be fine :) > El 20/07/2011, a las 11:54a.m., Igor Stasenko escribiÃ³: > >> I changed the MakeUUID in sqUnixUUID.c to look like following: >> >> >> >> int MakeUUID(char *location) >> { >> Â uuid_t uuid; >> >> #if defined(HAVE_UUIDGEN) >> Â uuidgen(&uuid, 1); >> #elif defined(HAVE_UUID_GENERATE) >> Â uuid_generate(uuid); >> #else >> #error "No uuid function provided" >> #endif >> >> Â memcpy((void *)location, (void *)&uuid, sizeof(uuid)); >> Â return 1; >> } >> >> >> and during compiling cocoa vms it fails with error. >> And before change the code was looking like following: >> >> int MakeUUID(char *location) >> { >> Â uuid_t uuid; >> >> #if defined(HAVE_UUIDGEN) >> Â uuidgen(&uuid, 1); >> #elif defined(HAVE_UUID_GENERATE) >> Â uuid_generate(uuid); >> #endif >> >> Â memcpy((void *)location, (void *)&uuid, sizeof(uuid)); >> Â return 1; >> } >> >> which means that if none HAVE_UUIDGEN or HAVE_UUID_GENERATE is >> defined, then there is nothing got called >> and memcpy then returns a random values on stack, copied from uuid to locaiton. >> >> And first version of that file just contains : >> >> >> int MakeUUID(char *location) >> { >> Â uuid_t uuid; >> Â uuid_generate(uuid); >> Â memcpy((void *)location, (void *)&uuid, sizeof(uuid)); >> Â return 1; >> } >> >> and then changed to: >> >> #if defined(__NetBSD__) >> Â uuidgen(&uuid, 1); >> #else >> Â uuid_generate(uuid); >> #endif >> >> and then changed to: >> >> #if defined(HAVE_UUIDGEN) >> Â uuidgen(&uuid, 1); >> #elif defined(HAVE_UUID_GENERATE) >> Â uuid_generate(uuid); >> #endif >> >> Blame: >> Merge Levente's linux UUIDPlugin fix. Fix tickerSleepCycle decl in >> heartbeat. Restore -O2 optimization for the interpreter. >> >> Author: Â Â Â eliot >> Date: 5 months ago >> Commit: Â Â Â 757297f62c32eaf2af7f26ce66aa6499df776749 >> >> >> >> -- >> Best regards, >> Igor Stasenko AKA sig. >> > > > -- Best regards, Igor Stasenko AKA sig.

July 20, 2011

Re: [Pharo-project] UUIDs not so unique

by Esteban Lorenzano

no idea :) maybe because historical reasons. I'll see that, and change if needed. cheers, Esteban El 20/07/2011, a las 12:01p.m., Igor Stasenko escribiÃ³: > Btw, Esteban > i wonder why you using sqUnixUUID.c in cocoa configs, > while MacOS config using sqMacUUID.c which using mac-specific function > to generate uuid. > See > Mac OS/plugins/UUIDPlugin/sqMacUUID.c > > and apparently don't have the problem with defines. > > -- > Best regards, > Igor Stasenko AKA sig. >

July 20, 2011