Re: Lisp filepath issues when running application directly from jar file
This "idiot" refactored a very small portion of your code making it MUCH faster, and also passed more than 90% of the tests ABCL is currently failing, leaving only 3 ansi-tests that are still failing. If I'm an idiot it doesn't say much about you since you've been working on this for years and I've only been working on it for weeks. Best Regards, Ralph Ritoch On Mon, Aug 17, 2015 at 8:00 PM, <armedbear-devel-request@common-lisp.net> wrote:
Send armedbear-devel mailing list submissions to armedbear-devel@common-lisp.net
To subscribe or unsubscribe via the World Wide Web, visit https://mailman.common-lisp.net/listinfo/armedbear-devel or, via email, send a message with subject or body 'help' to armedbear-devel-request@common-lisp.net
You can reach the person managing the list at armedbear-devel-owner@common-lisp.net
When replying, please edit your Subject line so it is more specific than "Re: Contents of armedbear-devel digest..."
Today's Topics:
1. Re: Lisp filepath issues when running application directly from jar file (Mark Evenson)
----------------------------------------------------------------------
Message: 1 Date: Sun, 16 Aug 2015 14:38:44 +0200 From: Mark Evenson <evenson@panix.com> To: Armed Bear <armedbear-devel@common-lisp.net> Subject: Re: Lisp filepath issues when running application directly from jar file Message-ID: <5616917D-F789-4F95-93E2-FC3B7633B051@panix.com> Content-Type: text/plain; charset="utf-8"
On Aug 16, 2015, at 02:34, Ralph Ritoch <rritoch@gmail.com> wrote:
I have used this function in my code without ANY problems. [?]
The SYSTEM package is not intended to be used by users unless there is no alternative. The manual states clearly that functionality provided by SYSTEM is subject to be changed between major releases without warning. Therefore, the argument in this case isn?t that one would have problems using SYSTEM:LOAD-SYSTEM-FILE, but rather that one should use the ANSI interface as implemented CL:LOAD which should be able to load Lisp and/or fasls in all cases needed by the user. My work on EXT:JAR-PATHNAME and EXT:URL-PATHNAME was motivated by precisely the need for CL:LOAD to work in all conceivable cases in the JVM ecology. And the best contemporary mechanism for expression for resolution of and dependency order for CL:LOAD lies in ASDF, hence my emphasis on using its abstraction where extending CL:LOAD is not desirable.
As a side note, over time, we wish to move functions in SYSTEM which useful to the user into the EXTENSION package, but this work is incomplete at best. Currently, there are indeed cases in which using a function from SYSTEM is unavoidable. The use of SYSTEM:LOAD-SYSTEM-FILE is not one of them.
Your attitude regarding "progress" is archaic. If you don't want to support maven or other java technologies why do you even bother controlling abcl?
Your attitude towards discussion is not only archaic (i.e. immature) but counterproductive towards reaching the sort of consensus needed for progress.
However misguided such an endeavor may be, a couple points follow towards a defense from your sleazy ad hominem ?argumentation?.
As for supporting Maven, I have developed the machinery to use Maven artifacts in ABCL-ASDF, and continued to maintain them (with help from users) even as the lurching horror that comprises Maven release engineering continues to change API semantics between patch level releases. I have offered to collaborate with you on incorporating your initial work on building from Maven, pointing out the problems that I see with your approach to which I have received no reply.
As for supporting ?other technologies?, I suppose you are referring to your attempt to merge abcl-contrib.jar and abcl.jar. In this effort, I pointed out you could do this cleanly with the current code by adding site-specific code to the ?system.lisp? file whose contents are controlled by the Ant build process. Since your Maven build doesn?t use the Ant target to package (one of my explicit criticisms), you presumedly weren?t interested in using this mechanism. I responded twice to your proposal to probe jar files on the classpath for code to pontential load with suggestions which you found too burdensome to discuss so you developed what you needed. Your email announcing this implementation implied you had implemented to a ?specification?, but no such entity existed external to the source code. If you wish to provide such a specification for your current implementation, please do so that others may consider its usefulness.
As to whether I control ABCL, I was going to reply that I am but one maintainer out of five; that you are welcome to convince a majority of the maintainers through useful and accepted contributions that that you should be in ?control? too; and that since the code is GPL?d so you are welcome to get the fork out of here. But at this point, I would reply that if I do somehow ?control? ABCL, I am glad to protect it from idiots like you.
-- "A screaming comes across the sky. It has happened before but there is nothing to compare to it now."
On Mon, Aug 17, 2015 at 2:04 PM, Ralph Ritoch <rritoch@gmail.com> wrote:
This "idiot" refactored a very small portion of your code making it MUCH faster, and also passed more than 90% of the tests ABCL is currently failing, leaving only 3 ansi-tests that are still failing. If I'm an idiot it doesn't say much about you since you've been working on this for years and I've only been working on it for weeks.
True. But it's not like you solved all these issues by fundamentally supporting what they're testing for. The complex types that you fixed (and I already said this in IRC) could have been fixed the way you did over 10 years ago. We find it more important to fix the underlying issues. I'd like to note here that compliance doesn't come from "not failing the ansi tests", but from implementing the features the tests are trying to test. In other words, you're not compliant with a class of issues by being able to produce on specific instance of the class. The fix to make things "MUCH faster" as you say doesn't solve problems in the bigger picture; does the ABCL in your git repository now have fewer CL-TEST-GRID failures? If not, then it's not more useful than it was before: it doesn't run more of the most-used CL code out there.... Now, *that* would have been a great and useful contribution. I'm affraid that coming rushing in, expecting immediate answers, solving unimportart problems, calling people names if they don't respond quickly enough by your standards and then hard-forking the project all don't classify as great contributions. I hoped that we could get the progress you said you wanted through the discussion that's to be expected when you are an open source contributor. Alas, as Mark points out, there hasn't been much of the discussion I had hoped for... Regards, Erik.
Best Regards, Ralph Ritoch
On Mon, Aug 17, 2015 at 8:00 PM, <armedbear-devel-request@common-lisp.net> wrote:
Send armedbear-devel mailing list submissions to armedbear-devel@common-lisp.net
To subscribe or unsubscribe via the World Wide Web, visit https://mailman.common-lisp.net/listinfo/armedbear-devel or, via email, send a message with subject or body 'help' to armedbear-devel-request@common-lisp.net
You can reach the person managing the list at armedbear-devel-owner@common-lisp.net
When replying, please edit your Subject line so it is more specific than "Re: Contents of armedbear-devel digest..."
Today's Topics:
1. Re: Lisp filepath issues when running application directly from jar file (Mark Evenson)
----------------------------------------------------------------------
Message: 1 Date: Sun, 16 Aug 2015 14:38:44 +0200 From: Mark Evenson <evenson@panix.com> To: Armed Bear <armedbear-devel@common-lisp.net> Subject: Re: Lisp filepath issues when running application directly from jar file Message-ID: <5616917D-F789-4F95-93E2-FC3B7633B051@panix.com> Content-Type: text/plain; charset="utf-8"
On Aug 16, 2015, at 02:34, Ralph Ritoch <rritoch@gmail.com> wrote:
I have used this function in my code without ANY problems. [?]
The SYSTEM package is not intended to be used by users unless there is no alternative. The manual states clearly that functionality provided by SYSTEM is subject to be changed between major releases without warning. Therefore, the argument in this case isn?t that one would have problems using SYSTEM:LOAD-SYSTEM-FILE, but rather that one should use the ANSI interface as implemented CL:LOAD which should be able to load Lisp and/or fasls in all cases needed by the user. My work on EXT:JAR-PATHNAME and EXT:URL-PATHNAME was motivated by precisely the need for CL:LOAD to work in all conceivable cases in the JVM ecology. And the best contemporary mechanism for expression for resolution of and dependency order for CL:LOAD lies in ASDF, hence my emphasis on using its abstraction where extending CL:LOAD is not desirable.
As a side note, over time, we wish to move functions in SYSTEM which useful to the user into the EXTENSION package, but this work is incomplete at best. Currently, there are indeed cases in which using a function from SYSTEM is unavoidable. The use of SYSTEM:LOAD-SYSTEM-FILE is not one of them.
Your attitude regarding "progress" is archaic. If you don't want to support maven or other java technologies why do you even bother controlling abcl?
Your attitude towards discussion is not only archaic (i.e. immature) but counterproductive towards reaching the sort of consensus needed for progress.
However misguided such an endeavor may be, a couple points follow towards a defense from your sleazy ad hominem ?argumentation?.
As for supporting Maven, I have developed the machinery to use Maven artifacts in ABCL-ASDF, and continued to maintain them (with help from users) even as the lurching horror that comprises Maven release engineering continues to change API semantics between patch level releases. I have offered to collaborate with you on incorporating your initial work on building from Maven, pointing out the problems that I see with your approach to which I have received no reply.
As for supporting ?other technologies?, I suppose you are referring to your attempt to merge abcl-contrib.jar and abcl.jar. In this effort, I pointed out you could do this cleanly with the current code by adding site-specific code to the ?system.lisp? file whose contents are controlled by the Ant build process. Since your Maven build doesn?t use the Ant target to package (one of my explicit criticisms), you presumedly weren?t interested in using this mechanism. I responded twice to your proposal to probe jar files on the classpath for code to pontential load with suggestions which you found too burdensome to discuss so you developed what you needed. Your email announcing this implementation implied you had implemented to a ?specification?, but no such entity existed external to the source code. If you wish to provide such a specification for your current implementation, please do so that others may consider its usefulness.
As to whether I control ABCL, I was going to reply that I am but one maintainer out of five; that you are welcome to convince a majority of the maintainers through useful and accepted contributions that that you should be in ?control? too; and that since the code is GPL?d so you are welcome to get the fork out of here. But at this point, I would reply that if I do somehow ?control? ABCL, I am glad to protect it from idiots like you.
-- "A screaming comes across the sky. It has happened before but there is nothing to compare to it now."
Erik, Your pointing out ONE fix that didn't completely solve ALL problems related to compound types, but that one fix didn't solve all of the failures. Most of the repairs I made were complete. I don't expect my fork to run most LISP applications because it is only being re-designed to support compliant applications. Specifically I dropped automatic support for the #No-Break_Space named character. ABCL does not meet my needs, period, and it doesn't meet the needs of most business applications. It won't be adopted by the industry until it is refactored to meet the needs of the industry, and that is what I am working on. As mark says, he doesn't want to allow users to have access to loading lisp from the classloader. Appropriate maven integration was also flat-out rejected, and his solution of tying system dependent configuration settings to the deployment of applications is not conducive to Interoperability. At the end of the day it all comes down to two questions, does ABCL meet my needs today, and will it meet those needs in the future. Mark has made it clear that the answers to both questions are a resounding NO. Anyone looking to include a LISP in an enterprise application is going to come to the same conclusion, leaving ABCL in a subculture that can never be part of the mainstream because it doesn't meet the needs of the industry. Best Regards, Ralph Ritoch On Mon, Aug 17, 2015 at 8:52 PM, Erik Huelsmann <ehuels@gmail.com> wrote:
On Mon, Aug 17, 2015 at 2:04 PM, Ralph Ritoch <rritoch@gmail.com> wrote:
This "idiot" refactored a very small portion of your code making it MUCH faster, and also passed more than 90% of the tests ABCL is currently failing, leaving only 3 ansi-tests that are still failing. If I'm an idiot it doesn't say much about you since you've been working on this for years and I've only been working on it for weeks.
True. But it's not like you solved all these issues by fundamentally supporting what they're testing for. The complex types that you fixed (and I already said this in IRC) could have been fixed the way you did over 10 years ago. We find it more important to fix the underlying issues. I'd like to note here that compliance doesn't come from "not failing the ansi tests", but from implementing the features the tests are trying to test. In other words, you're not compliant with a class of issues by being able to produce on specific instance of the class.
The fix to make things "MUCH faster" as you say doesn't solve problems in the bigger picture; does the ABCL in your git repository now have fewer CL-TEST-GRID failures? If not, then it's not more useful than it was before: it doesn't run more of the most-used CL code out there.... Now, *that* would have been a great and useful contribution. I'm affraid that coming rushing in, expecting immediate answers, solving unimportart problems, calling people names if they don't respond quickly enough by your standards and then hard-forking the project all don't classify as great contributions.
I hoped that we could get the progress you said you wanted through the discussion that's to be expected when you are an open source contributor. Alas, as Mark points out, there hasn't been much of the discussion I had hoped for...
Regards,
Erik.
Best Regards, Ralph Ritoch
On Mon, Aug 17, 2015 at 8:00 PM, <armedbear-devel-request@common-lisp.net
wrote:
Send armedbear-devel mailing list submissions to armedbear-devel@common-lisp.net
To subscribe or unsubscribe via the World Wide Web, visit https://mailman.common-lisp.net/listinfo/armedbear-devel or, via email, send a message with subject or body 'help' to armedbear-devel-request@common-lisp.net
You can reach the person managing the list at armedbear-devel-owner@common-lisp.net
When replying, please edit your Subject line so it is more specific than "Re: Contents of armedbear-devel digest..."
Today's Topics:
1. Re: Lisp filepath issues when running application directly from jar file (Mark Evenson)
----------------------------------------------------------------------
Message: 1 Date: Sun, 16 Aug 2015 14:38:44 +0200 From: Mark Evenson <evenson@panix.com> To: Armed Bear <armedbear-devel@common-lisp.net> Subject: Re: Lisp filepath issues when running application directly from jar file Message-ID: <5616917D-F789-4F95-93E2-FC3B7633B051@panix.com> Content-Type: text/plain; charset="utf-8"
On Aug 16, 2015, at 02:34, Ralph Ritoch <rritoch@gmail.com> wrote:
I have used this function in my code without ANY problems. [?]
The SYSTEM package is not intended to be used by users unless there is no alternative. The manual states clearly that functionality provided by SYSTEM is subject to be changed between major releases without warning. Therefore, the argument in this case isn?t that one would have problems using SYSTEM:LOAD-SYSTEM-FILE, but rather that one should use the ANSI interface as implemented CL:LOAD which should be able to load Lisp and/or fasls in all cases needed by the user. My work on EXT:JAR-PATHNAME and EXT:URL-PATHNAME was motivated by precisely the need for CL:LOAD to work in all conceivable cases in the JVM ecology. And the best contemporary mechanism for expression for resolution of and dependency order for CL:LOAD lies in ASDF, hence my emphasis on using its abstraction where extending CL:LOAD is not desirable.
As a side note, over time, we wish to move functions in SYSTEM which useful to the user into the EXTENSION package, but this work is incomplete at best. Currently, there are indeed cases in which using a function from SYSTEM is unavoidable. The use of SYSTEM:LOAD-SYSTEM-FILE is not one of them.
Your attitude regarding "progress" is archaic. If you don't want to support maven or other java technologies why do you even bother controlling abcl?
Your attitude towards discussion is not only archaic (i.e. immature) but counterproductive towards reaching the sort of consensus needed for progress.
However misguided such an endeavor may be, a couple points follow towards a defense from your sleazy ad hominem ?argumentation?.
As for supporting Maven, I have developed the machinery to use Maven artifacts in ABCL-ASDF, and continued to maintain them (with help from users) even as the lurching horror that comprises Maven release engineering continues to change API semantics between patch level releases. I have offered to collaborate with you on incorporating your initial work on building from Maven, pointing out the problems that I see with your approach to which I have received no reply.
As for supporting ?other technologies?, I suppose you are referring to your attempt to merge abcl-contrib.jar and abcl.jar. In this effort, I pointed out you could do this cleanly with the current code by adding site-specific code to the ?system.lisp? file whose contents are controlled by the Ant build process. Since your Maven build doesn?t use the Ant target to package (one of my explicit criticisms), you presumedly weren?t interested in using this mechanism. I responded twice to your proposal to probe jar files on the classpath for code to pontential load with suggestions which you found too burdensome to discuss so you developed what you needed. Your email announcing this implementation implied you had implemented to a ?specification?, but no such entity existed external to the source code. If you wish to provide such a specification for your current implementation, please do so that others may consider its usefulness.
As to whether I control ABCL, I was going to reply that I am but one maintainer out of five; that you are welcome to convince a majority of the maintainers through useful and accepted contributions that that you should be in ?control? too; and that since the code is GPL?d so you are welcome to get the fork out of here. But at this point, I would reply that if I do somehow ?control? ABCL, I am glad to protect it from idiots like you.
-- "A screaming comes across the sky. It has happened before but there is nothing to compare to it now."
Please, can anyone help me here? On Mon, Aug 17, 2015 at 2:03 PM, Ralph Ritoch <rritoch@gmail.com> wrote:
Erik,
Your pointing out ONE fix that didn't completely solve ALL problems related to compound types, but that one fix didn't solve all of the failures. Most of the repairs I made were complete. I don't expect my fork to run most LISP applications because it is only being re-designed to support compliant applications. Specifically I dropped automatic support for the #No-Break_Space named character. ABCL does not meet my needs, period, and it doesn't meet the needs of most business applications. It won't be adopted by the industry until it is refactored to meet the needs of the industry, and that is what I am working on. As mark says, he doesn't want to allow users to have access to loading lisp from the classloader. Appropriate maven integration was also flat-out rejected, and his solution of tying system dependent configuration settings to the deployment of applications is not conducive to Interoperability. At the end of the day it all comes down to two questions, does ABCL meet my needs today, and will it meet those needs in the future. Mark has made it clear that the answers to both questions are a resounding NO. Anyone looking to include a LISP in an enterprise application is going to come to the same conclusion, leaving ABCL in a subculture that can never be part of the mainstream because it doesn't meet the needs of the industry.
Best Regards, Ralph Ritoch
On Mon, Aug 17, 2015 at 8:52 PM, Erik Huelsmann <ehuels@gmail.com> wrote:
On Mon, Aug 17, 2015 at 2:04 PM, Ralph Ritoch <rritoch@gmail.com> wrote:
This "idiot" refactored a very small portion of your code making it MUCH faster, and also passed more than 90% of the tests ABCL is currently failing, leaving only 3 ansi-tests that are still failing. If I'm an idiot it doesn't say much about you since you've been working on this for years and I've only been working on it for weeks.
True. But it's not like you solved all these issues by fundamentally supporting what they're testing for. The complex types that you fixed (and I already said this in IRC) could have been fixed the way you did over 10 years ago. We find it more important to fix the underlying issues. I'd like to note here that compliance doesn't come from "not failing the ansi tests", but from implementing the features the tests are trying to test. In other words, you're not compliant with a class of issues by being able to produce on specific instance of the class.
The fix to make things "MUCH faster" as you say doesn't solve problems in the bigger picture; does the ABCL in your git repository now have fewer CL-TEST-GRID failures? If not, then it's not more useful than it was before: it doesn't run more of the most-used CL code out there.... Now, *that* would have been a great and useful contribution. I'm affraid that coming rushing in, expecting immediate answers, solving unimportart problems, calling people names if they don't respond quickly enough by your standards and then hard-forking the project all don't classify as great contributions.
I hoped that we could get the progress you said you wanted through the discussion that's to be expected when you are an open source contributor. Alas, as Mark points out, there hasn't been much of the discussion I had hoped for...
Regards,
Erik.
Best Regards, Ralph Ritoch
On Mon, Aug 17, 2015 at 8:00 PM, < armedbear-devel-request@common-lisp.net> wrote:
Send armedbear-devel mailing list submissions to armedbear-devel@common-lisp.net
To subscribe or unsubscribe via the World Wide Web, visit https://mailman.common-lisp.net/listinfo/armedbear-devel or, via email, send a message with subject or body 'help' to armedbear-devel-request@common-lisp.net
You can reach the person managing the list at armedbear-devel-owner@common-lisp.net
When replying, please edit your Subject line so it is more specific than "Re: Contents of armedbear-devel digest..."
Today's Topics:
1. Re: Lisp filepath issues when running application directly from jar file (Mark Evenson)
----------------------------------------------------------------------
Message: 1 Date: Sun, 16 Aug 2015 14:38:44 +0200 From: Mark Evenson <evenson@panix.com> To: Armed Bear <armedbear-devel@common-lisp.net> Subject: Re: Lisp filepath issues when running application directly from jar file Message-ID: <5616917D-F789-4F95-93E2-FC3B7633B051@panix.com> Content-Type: text/plain; charset="utf-8"
On Aug 16, 2015, at 02:34, Ralph Ritoch <rritoch@gmail.com> wrote:
I have used this function in my code without ANY problems. [?]
The SYSTEM package is not intended to be used by users unless there is no alternative. The manual states clearly that functionality provided by SYSTEM is subject to be changed between major releases without warning. Therefore, the argument in this case isn?t that one would have problems using SYSTEM:LOAD-SYSTEM-FILE, but rather that one should use the ANSI interface as implemented CL:LOAD which should be able to load Lisp and/or fasls in all cases needed by the user. My work on EXT:JAR-PATHNAME and EXT:URL-PATHNAME was motivated by precisely the need for CL:LOAD to work in all conceivable cases in the JVM ecology. And the best contemporary mechanism for expression for resolution of and dependency order for CL:LOAD lies in ASDF, hence my emphasis on using its abstraction where extending CL:LOAD is not desirable.
As a side note, over time, we wish to move functions in SYSTEM which useful to the user into the EXTENSION package, but this work is incomplete at best. Currently, there are indeed cases in which using a function from SYSTEM is unavoidable. The use of SYSTEM:LOAD-SYSTEM-FILE is not one of them.
Your attitude regarding "progress" is archaic. If you don't want to support maven or other java technologies why do you even bother controlling abcl?
Your attitude towards discussion is not only archaic (i.e. immature) but counterproductive towards reaching the sort of consensus needed for progress.
However misguided such an endeavor may be, a couple points follow towards a defense from your sleazy ad hominem ?argumentation?.
As for supporting Maven, I have developed the machinery to use Maven artifacts in ABCL-ASDF, and continued to maintain them (with help from users) even as the lurching horror that comprises Maven release engineering continues to change API semantics between patch level releases. I have offered to collaborate with you on incorporating your initial work on building from Maven, pointing out the problems that I see with your approach to which I have received no reply.
As for supporting ?other technologies?, I suppose you are referring to your attempt to merge abcl-contrib.jar and abcl.jar. In this effort, I pointed out you could do this cleanly with the current code by adding site-specific code to the ?system.lisp? file whose contents are controlled by the Ant build process. Since your Maven build doesn?t use the Ant target to package (one of my explicit criticisms), you presumedly weren?t interested in using this mechanism. I responded twice to your proposal to probe jar files on the classpath for code to pontential load with suggestions which you found too burdensome to discuss so you developed what you needed. Your email announcing this implementation implied you had implemented to a ?specification?, but no such entity existed external to the source code. If you wish to provide such a specification for your current implementation, please do so that others may consider its usefulness.
As to whether I control ABCL, I was going to reply that I am but one maintainer out of five; that you are welcome to convince a majority of the maintainers through useful and accepted contributions that that you should be in ?control? too; and that since the code is GPL?d so you are welcome to get the fork out of here. But at this point, I would reply that if I do somehow ?control? ABCL, I am glad to protect it from idiots like you.
-- "A screaming comes across the sky. It has happened before but there is nothing to compare to it now."
On 17 Aug 2015, at 15:04, Hamda Binte Ajmal <hamda.binte.ajmal@gmail.com> wrote:
Please, can anyone help me here?
Probably. But: 1- post a new message, not an answer in a thread that is unrelated to your problem. (by now, almost everybody is skipping this thread). 2- put a summary of your problem in the subject: line. 3- explain your problem in the body of the message, saying what you tried, what you got and what you expected. -- __Pascal J. Bourguignon__
Hamda Binte Ajmal, From ABCL/Lisp what you want to do, loading from the classloader, is NOT possible unless your willing to use (load-system-file "path/to/resource.lisp") which is provided by the system package. Note the defpackage I used to access the system package. https://github.com/rritoch/abclp/blob/master/src/main/abclp/META-INF/abcl/ab...
From Java side you can call Load.loadSystemFile as you can see here.
https://github.com/rritoch/abclp/blob/master/src/main/java/com/vnetpublishin... Mark makes it very clear that he doesn't want these features available to users so there is a high possibility that they will add additional code to make it more difficult to access this capability. As I stated in my previous email, ABCL just doesn't meet the needs of our applications, but Mark doesn't seem to care. Best Regards, Ralph Ritoch On Mon, Aug 17, 2015 at 9:04 PM, Hamda Binte Ajmal < hamda.binte.ajmal@gmail.com> wrote:
Please, can anyone help me here?
On Mon, Aug 17, 2015 at 2:03 PM, Ralph Ritoch <rritoch@gmail.com> wrote:
Erik,
Your pointing out ONE fix that didn't completely solve ALL problems related to compound types, but that one fix didn't solve all of the failures. Most of the repairs I made were complete. I don't expect my fork to run most LISP applications because it is only being re-designed to support compliant applications. Specifically I dropped automatic support for the #No-Break_Space named character. ABCL does not meet my needs, period, and it doesn't meet the needs of most business applications. It won't be adopted by the industry until it is refactored to meet the needs of the industry, and that is what I am working on. As mark says, he doesn't want to allow users to have access to loading lisp from the classloader. Appropriate maven integration was also flat-out rejected, and his solution of tying system dependent configuration settings to the deployment of applications is not conducive to Interoperability. At the end of the day it all comes down to two questions, does ABCL meet my needs today, and will it meet those needs in the future. Mark has made it clear that the answers to both questions are a resounding NO. Anyone looking to include a LISP in an enterprise application is going to come to the same conclusion, leaving ABCL in a subculture that can never be part of the mainstream because it doesn't meet the needs of the industry.
Best Regards, Ralph Ritoch
On Mon, Aug 17, 2015 at 8:52 PM, Erik Huelsmann <ehuels@gmail.com> wrote:
On Mon, Aug 17, 2015 at 2:04 PM, Ralph Ritoch <rritoch@gmail.com> wrote:
This "idiot" refactored a very small portion of your code making it MUCH faster, and also passed more than 90% of the tests ABCL is currently failing, leaving only 3 ansi-tests that are still failing. If I'm an idiot it doesn't say much about you since you've been working on this for years and I've only been working on it for weeks.
True. But it's not like you solved all these issues by fundamentally supporting what they're testing for. The complex types that you fixed (and I already said this in IRC) could have been fixed the way you did over 10 years ago. We find it more important to fix the underlying issues. I'd like to note here that compliance doesn't come from "not failing the ansi tests", but from implementing the features the tests are trying to test. In other words, you're not compliant with a class of issues by being able to produce on specific instance of the class.
The fix to make things "MUCH faster" as you say doesn't solve problems in the bigger picture; does the ABCL in your git repository now have fewer CL-TEST-GRID failures? If not, then it's not more useful than it was before: it doesn't run more of the most-used CL code out there.... Now, *that* would have been a great and useful contribution. I'm affraid that coming rushing in, expecting immediate answers, solving unimportart problems, calling people names if they don't respond quickly enough by your standards and then hard-forking the project all don't classify as great contributions.
I hoped that we could get the progress you said you wanted through the discussion that's to be expected when you are an open source contributor. Alas, as Mark points out, there hasn't been much of the discussion I had hoped for...
Regards,
Erik.
Best Regards, Ralph Ritoch
On Mon, Aug 17, 2015 at 8:00 PM, < armedbear-devel-request@common-lisp.net> wrote:
Send armedbear-devel mailing list submissions to armedbear-devel@common-lisp.net
To subscribe or unsubscribe via the World Wide Web, visit https://mailman.common-lisp.net/listinfo/armedbear-devel or, via email, send a message with subject or body 'help' to armedbear-devel-request@common-lisp.net
You can reach the person managing the list at armedbear-devel-owner@common-lisp.net
When replying, please edit your Subject line so it is more specific than "Re: Contents of armedbear-devel digest..."
Today's Topics:
1. Re: Lisp filepath issues when running application directly from jar file (Mark Evenson)
----------------------------------------------------------------------
Message: 1 Date: Sun, 16 Aug 2015 14:38:44 +0200 From: Mark Evenson <evenson@panix.com> To: Armed Bear <armedbear-devel@common-lisp.net> Subject: Re: Lisp filepath issues when running application directly from jar file Message-ID: <5616917D-F789-4F95-93E2-FC3B7633B051@panix.com> Content-Type: text/plain; charset="utf-8"
On Aug 16, 2015, at 02:34, Ralph Ritoch <rritoch@gmail.com> wrote:
I have used this function in my code without ANY problems. [?]
The SYSTEM package is not intended to be used by users unless there is no alternative. The manual states clearly that functionality provided by SYSTEM is subject to be changed between major releases without warning. Therefore, the argument in this case isn?t that one would have problems using SYSTEM:LOAD-SYSTEM-FILE, but rather that one should use the ANSI interface as implemented CL:LOAD which should be able to load Lisp and/or fasls in all cases needed by the user. My work on EXT:JAR-PATHNAME and EXT:URL-PATHNAME was motivated by precisely the need for CL:LOAD to work in all conceivable cases in the JVM ecology. And the best contemporary mechanism for expression for resolution of and dependency order for CL:LOAD lies in ASDF, hence my emphasis on using its abstraction where extending CL:LOAD is not desirable.
As a side note, over time, we wish to move functions in SYSTEM which useful to the user into the EXTENSION package, but this work is incomplete at best. Currently, there are indeed cases in which using a function from SYSTEM is unavoidable. The use of SYSTEM:LOAD-SYSTEM-FILE is not one of them.
Your attitude regarding "progress" is archaic. If you don't want to support maven or other java technologies why do you even bother controlling abcl?
Your attitude towards discussion is not only archaic (i.e. immature) but counterproductive towards reaching the sort of consensus needed for progress.
However misguided such an endeavor may be, a couple points follow towards a defense from your sleazy ad hominem ?argumentation?.
As for supporting Maven, I have developed the machinery to use Maven artifacts in ABCL-ASDF, and continued to maintain them (with help from users) even as the lurching horror that comprises Maven release engineering continues to change API semantics between patch level releases. I have offered to collaborate with you on incorporating your initial work on building from Maven, pointing out the problems that I see with your approach to which I have received no reply.
As for supporting ?other technologies?, I suppose you are referring to your attempt to merge abcl-contrib.jar and abcl.jar. In this effort, I pointed out you could do this cleanly with the current code by adding site-specific code to the ?system.lisp? file whose contents are controlled by the Ant build process. Since your Maven build doesn?t use the Ant target to package (one of my explicit criticisms), you presumedly weren?t interested in using this mechanism. I responded twice to your proposal to probe jar files on the classpath for code to pontential load with suggestions which you found too burdensome to discuss so you developed what you needed. Your email announcing this implementation implied you had implemented to a ?specification?, but no such entity existed external to the source code. If you wish to provide such a specification for your current implementation, please do so that others may consider its usefulness.
As to whether I control ABCL, I was going to reply that I am but one maintainer out of five; that you are welcome to convince a majority of the maintainers through useful and accepted contributions that that you should be in ?control? too; and that since the code is GPL?d so you are welcome to get the fork out of here. But at this point, I would reply that if I do somehow ?control? ABCL, I am glad to protect it from idiots like you.
-- "A screaming comes across the sky. It has happened before but there is nothing to compare to it now."
Hi Ralph, Pointing out what Mark doesn't want is only going to fuel the controversy and definitely not helping Hamda. Please stop doing that. Also please stop bashing ABCL on ABCL's mailing list. People are here because they are interested in ABCL and using ABCL. They are not interested in being told that ABCL is insufficient in some way or another that may or may not coincide with their goals. If ABCL turns out to be insufficient for people, they can come to that conclusion on their own with the *help* of the mailing list, working through the issues that they are experiencing. So, if you have anything to add to the discussion within the bounds of ABCL's *public api*, please do so; if not, please refrain from making unrelated comments and/or fuelling controversy. Regards, Erik. On Mon, Aug 17, 2015 at 3:14 PM, Ralph Ritoch <rritoch@gmail.com> wrote:
Hamda Binte Ajmal,
From ABCL/Lisp what you want to do, loading from the classloader, is NOT possible unless your willing to use (load-system-file "path/to/resource.lisp") which is provided by the system package.
Note the defpackage I used to access the system package.
https://github.com/rritoch/abclp/blob/master/src/main/abclp/META-INF/abcl/ab...
From Java side you can call Load.loadSystemFile as you can see here.
https://github.com/rritoch/abclp/blob/master/src/main/java/com/vnetpublishin...
Mark makes it very clear that he doesn't want these features available to users so there is a high possibility that they will add additional code to make it more difficult to access this capability. As I stated in my previous email, ABCL just doesn't meet the needs of our applications, but Mark doesn't seem to care.
Best Regards, Ralph Ritoch
On Mon, Aug 17, 2015 at 9:04 PM, Hamda Binte Ajmal < hamda.binte.ajmal@gmail.com> wrote:
Please, can anyone help me here?
On Mon, Aug 17, 2015 at 2:03 PM, Ralph Ritoch <rritoch@gmail.com> wrote:
Erik,
Your pointing out ONE fix that didn't completely solve ALL problems related to compound types, but that one fix didn't solve all of the failures. Most of the repairs I made were complete. I don't expect my fork to run most LISP applications because it is only being re-designed to support compliant applications. Specifically I dropped automatic support for the #No-Break_Space named character. ABCL does not meet my needs, period, and it doesn't meet the needs of most business applications. It won't be adopted by the industry until it is refactored to meet the needs of the industry, and that is what I am working on. As mark says, he doesn't want to allow users to have access to loading lisp from the classloader. Appropriate maven integration was also flat-out rejected, and his solution of tying system dependent configuration settings to the deployment of applications is not conducive to Interoperability. At the end of the day it all comes down to two questions, does ABCL meet my needs today, and will it meet those needs in the future. Mark has made it clear that the answers to both questions are a resounding NO. Anyone looking to include a LISP in an enterprise application is going to come to the same conclusion, leaving ABCL in a subculture that can never be part of the mainstream because it doesn't meet the needs of the industry.
Best Regards, Ralph Ritoch
On Mon, Aug 17, 2015 at 8:52 PM, Erik Huelsmann <ehuels@gmail.com> wrote:
On Mon, Aug 17, 2015 at 2:04 PM, Ralph Ritoch <rritoch@gmail.com> wrote:
This "idiot" refactored a very small portion of your code making it MUCH faster, and also passed more than 90% of the tests ABCL is currently failing, leaving only 3 ansi-tests that are still failing. If I'm an idiot it doesn't say much about you since you've been working on this for years and I've only been working on it for weeks.
True. But it's not like you solved all these issues by fundamentally supporting what they're testing for. The complex types that you fixed (and I already said this in IRC) could have been fixed the way you did over 10 years ago. We find it more important to fix the underlying issues. I'd like to note here that compliance doesn't come from "not failing the ansi tests", but from implementing the features the tests are trying to test. In other words, you're not compliant with a class of issues by being able to produce on specific instance of the class.
The fix to make things "MUCH faster" as you say doesn't solve problems in the bigger picture; does the ABCL in your git repository now have fewer CL-TEST-GRID failures? If not, then it's not more useful than it was before: it doesn't run more of the most-used CL code out there.... Now, *that* would have been a great and useful contribution. I'm affraid that coming rushing in, expecting immediate answers, solving unimportart problems, calling people names if they don't respond quickly enough by your standards and then hard-forking the project all don't classify as great contributions.
I hoped that we could get the progress you said you wanted through the discussion that's to be expected when you are an open source contributor. Alas, as Mark points out, there hasn't been much of the discussion I had hoped for...
Regards,
Erik.
Best Regards, Ralph Ritoch
On Mon, Aug 17, 2015 at 8:00 PM, < armedbear-devel-request@common-lisp.net> wrote:
Send armedbear-devel mailing list submissions to armedbear-devel@common-lisp.net
To subscribe or unsubscribe via the World Wide Web, visit https://mailman.common-lisp.net/listinfo/armedbear-devel or, via email, send a message with subject or body 'help' to armedbear-devel-request@common-lisp.net
You can reach the person managing the list at armedbear-devel-owner@common-lisp.net
When replying, please edit your Subject line so it is more specific than "Re: Contents of armedbear-devel digest..."
Today's Topics:
1. Re: Lisp filepath issues when running application directly from jar file (Mark Evenson)
----------------------------------------------------------------------
Message: 1 Date: Sun, 16 Aug 2015 14:38:44 +0200 From: Mark Evenson <evenson@panix.com> To: Armed Bear <armedbear-devel@common-lisp.net> Subject: Re: Lisp filepath issues when running application directly from jar file Message-ID: <5616917D-F789-4F95-93E2-FC3B7633B051@panix.com> Content-Type: text/plain; charset="utf-8"
> On Aug 16, 2015, at 02:34, Ralph Ritoch <rritoch@gmail.com> wrote: > > I have used this function in my code without ANY problems. [?]
The SYSTEM package is not intended to be used by users unless there is no alternative. The manual states clearly that functionality provided by SYSTEM is subject to be changed between major releases without warning. Therefore, the argument in this case isn?t that one would have problems using SYSTEM:LOAD-SYSTEM-FILE, but rather that one should use the ANSI interface as implemented CL:LOAD which should be able to load Lisp and/or fasls in all cases needed by the user. My work on EXT:JAR-PATHNAME and EXT:URL-PATHNAME was motivated by precisely the need for CL:LOAD to work in all conceivable cases in the JVM ecology. And the best contemporary mechanism for expression for resolution of and dependency order for CL:LOAD lies in ASDF, hence my emphasis on using its abstraction where extending CL:LOAD is not desirable.
As a side note, over time, we wish to move functions in SYSTEM which useful to the user into the EXTENSION package, but this work is incomplete at best. Currently, there are indeed cases in which using a function from SYSTEM is unavoidable. The use of SYSTEM:LOAD-SYSTEM-FILE is not one of them.
> Your attitude regarding "progress" is archaic. If you don't want to support maven or other java technologies why do you even bother controlling abcl?
Your attitude towards discussion is not only archaic (i.e. immature) but counterproductive towards reaching the sort of consensus needed for progress.
However misguided such an endeavor may be, a couple points follow towards a defense from your sleazy ad hominem ?argumentation?.
As for supporting Maven, I have developed the machinery to use Maven artifacts in ABCL-ASDF, and continued to maintain them (with help from users) even as the lurching horror that comprises Maven release engineering continues to change API semantics between patch level releases. I have offered to collaborate with you on incorporating your initial work on building from Maven, pointing out the problems that I see with your approach to which I have received no reply.
As for supporting ?other technologies?, I suppose you are referring to your attempt to merge abcl-contrib.jar and abcl.jar. In this effort, I pointed out you could do this cleanly with the current code by adding site-specific code to the ?system.lisp? file whose contents are controlled by the Ant build process. Since your Maven build doesn?t use the Ant target to package (one of my explicit criticisms), you presumedly weren?t interested in using this mechanism. I responded twice to your proposal to probe jar files on the classpath for code to pontential load with suggestions which you found too burdensome to discuss so you developed what you needed. Your email announcing this implementation implied you had implemented to a ?specification?, but no such entity existed external to the source code. If you wish to provide such a specification for your current implementation, please do so that others may consider its usefulness.
As to whether I control ABCL, I was going to reply that I am but one maintainer out of five; that you are welcome to convince a majority of the maintainers through useful and accepted contributions that that you should be in ?control? too; and that since the code is GPL?d so you are welcome to get the fork out of here. But at this point, I would reply that if I do somehow ?control? ABCL, I am glad to protect it from idiots like you.
-- "A screaming comes across the sky. It has happened before but there is nothing to compare to it now."
On Mon, Aug 17, 2015 at 6:23 AM, Erik Huelsmann <ehuels@gmail.com> wrote:
If ABCL turns out to be insufficient for people, they can come to that conclusion on their own with the *help* of the mailing list, working through the issues that they are experiencing.
Mark isn't just writing polite words. My experience, at least, worked exactly like that. I haven't been here in a while, because abcl just works for me and has for a long time. But there was a time (~5 years ago) when abcl *didn't* work for me. This list was incredibly helpful during that time, and helped me transform naive suggestions into things that worked for both me and abcl. The process wasn't entirely easy: there were problems with both what I wanted to do *and* with abcl, but the list (and numerous off list exchanges) helped me work things out. Since then my application has sat in it's (virtual) corner just doing it's job day after day, and I have had a need to really push abcl. But the experience was so positive that I still use abcl in situations it might not be ideal for, just because I like it and this community so much. Obviously, everybody will have different needs and experiences. I apologize for keeping this thread alive, and I certainly don't want to discount any body else's experience with the maintainers. But as I read this I thought it's a bit sad that they only hear from outsiders when things aren't working! -david
participants (5)
-
David Kirkman -
Erik Huelsmann -
Hamda Binte Ajmal -
Pascal J. Bourguignon -
Ralph Ritoch