Re: RFR 8009517: Disable fatal compiler warning in the old build
On 21/03/2013 22:12, Brad Wetmore wrote: : The codereview is here: http://cr.openjdk.java.net/~wetmore/8009517/webrev.00/ I plan to push through the deploy gate, as they have an integration next week. Thomas Ng will do the push for us. Any objections, please speak now. No objection here but just to mention that since you need to set NEWBUILD=false then it might not be too much extra to also set JAVAC_WARNINGS_FATAL=false. In any case, I think it would be desirable if there was a retirement date set for the old build so that the remaining users (I assume very few at this point) have something to aim for. -Alan.
Re: RFR 8009517: Disable fatal compiler warning in the old build
Brad, I do not build using the old build anymore. This is clearly a blocker for your work. If you want to suppress the warnings for overrides/deprecation, then please push the change ( your patch ). We can revisit this in the future, when it is necessary. -Chris. On 03/19/2013 01:29 AM, Brad Wetmore wrote: Sorry for the delay in response, I've been pulled in yet another direction, and this has come back up in priority. On 3/9/2013 12:11 AM, Chris Hegarty wrote: I agree about warning creeping problems. This is a temporary solution, we should soon be fixing the underlying hashcode/equals problems...but... Your temporary solution, -overrides, is just that. It will enable the old build to complete today, but it could fail at any point in the future, as the code changes. Correct. As it stands today, a recent change now requires *BOTH* overrides/deprecation in order to get a complete MASTER build using the old build system. [brwetmor@flicker-vm1] 222 hg diff common/shared/Defs-java.gmk diff --git a/make/common/shared/Defs-java.gmk b/make/common/shared/Defs-java.gmk --- a/make/common/shared/Defs-java.gmk +++ b/make/common/shared/Defs-java.gmk @@ -127,7 +127,7 @@ endif # TODO: Workaround for CR 7063027. Remove -path eventually. -JAVAC_LINT_OPTIONS += -Xlint:-path +JAVAC_LINT_OPTIONS += -Xlint:-path,-overrides,-deprecation JAVACFLAGS += $(JAVAC_LINT_OPTIONS) For example, java.net is currently warning free, in the old it compiles with fatal warnings enabled. Lets say, in a moment of madness, I add a dependency from java.net.Socket to say java.awt.RenderingHints.Key ( or any class that produces warnings when compiled. I run the new build, all is fine. Push the changes. Now someone else sync's up, but need to build using the old build. If the new dependent class is not already compiled before java.net.Socket gets compiled, it will be compiled implicitly. It's warnings will cause the compile to fail, and the old build will fail. Or much simpler, anyone could write sloppy code with warnings, the new build will suppress them, and they won't notice. Push this code, and the old build will fail if is explicitly, or implicitly, compiles this code with -Werror enabled. Exactly. Our formerly clean code now requires disabling of two Lint options, but the new build is happy just to report the warning. The old build crashes on the warning. Our options for the old build system are: 1. disable the warning for overrides/deprecation, keep -Werror (my preferred since these are minor warnings.) 2. Somehow disable -Werror on these new directories that are now failing. (more work to figure out, but also acceptable) 3. Fix the warnings. (I don't have cycles to drive a rewrite of use of deprecated code and/or add missing equals/hashcode that the recent javac changes exposed.) We spent a lot of time cleaning up many directories, seems a shame to start allowing non-fatal warnings to come back into previously clean code because people aren't taking the time to fix new warnings as they are introduced. I personally spent several weeks over the past number of years fixing warnings and reviewing warning cleanup webrevs from others. I took much pride in keeping certain areas warnings free. It is with great regret that I propose to disable fatal warnings in the old build, but I felt this the best/safest option. I heard much annoyance and frustration from others about hitting seemingly random errors with the old build recently. This is the only sure way to avoid that. The new builds will still warn, but the old builds will still fail for all but these override problems. Yes, you lose the warnings in the old, but seems better than completely shutting off erroring. I'm ok with that, if others are. To clarify, I think you are suggesting that we keep the old build as it, with -overrides, and now ,-deprecation :( and use it periodically as a way of tracking new warnings being introduced into areas that were warning free. That would be a side-effect, as someone would occasionally need to figure out what's changed. The main issue we're hitting right now is that RE has to make several source code changes in order to build JCE jar files without errors. I was able to change the individual LINT options globally and reduce it down to one change, but that's still one change that RE has to make. I feel that RE should not be making any changes, but that ship has already sailed and we're stuck with the results now. That is, if the old build fails because of a fatal warning, so be it. File a bug and fix the source code. Then the old build will work again. This means that at any point in time the old build cannot be guaranteed to be buildable. Everyone seems to agree, a solution needs to be found to allow us to keep certain areas warning free. This issue is too important, and too much time was spent, to allow it to regress to the state it was in a few years ago. It's already started. Brad
Re: RFR 8009517: Disable fatal compiler warning in the old build
I have a suggestion for how to at least partly enable -Werror in the new build. The penalty is slightly longer compile time, but the difference should be negligible. We split the big java compilation in jdk in two. The first pass with -Werror and all warnings turned on, the second without. We make a list of packages that are passing -Werror and use as include list for the first and exclude list for the second. As you make more packages warning free, we add them to the list. This solution is not as fine grained as a per package configured set of warning flags, but it's much better than we have today. /Erik On 2013-03-08 16:56, Alan Bateman wrote: On 08/03/2013 15:49, Mike Duigou wrote: Looks fine to me. Do we have an issue open for restoring warnings to the new build? Mike I don't know if there is an issue for that yet but as the new build compiles thousands of classes in a single compilation unit then it means we will need to make significant inroads on the warnings before more can be enabled. The approach with the old build was by area and good progress had been made but with the new build, then it may have to be by warning type as all areas are compiled together. -Alan.
Re: RFR 8009517: Disable fatal compiler warning in the old build
I tried implementing a PoC for this. Without sjavac, it works, except that the first pass must be run without -Werror and the second with. Since we use -implicit:none, this is fine. With sjavac I had to let it compile the full set of classes first and then run a second time (into a different output directory) with just the warning free set of packages. The overhead on my machine was 5 seconds for the second pass. This solution also works without sjavac, but then the overhead is 37 seconds on my machine. Now the question is, do we want to pursue this or not? /Erik On 2013-03-11 10:30, Erik Joelsson wrote: I have a suggestion for how to at least partly enable -Werror in the new build. The penalty is slightly longer compile time, but the difference should be negligible. We split the big java compilation in jdk in two. The first pass with -Werror and all warnings turned on, the second without. We make a list of packages that are passing -Werror and use as include list for the first and exclude list for the second. As you make more packages warning free, we add them to the list. This solution is not as fine grained as a per package configured set of warning flags, but it's much better than we have today. /Erik On 2013-03-08 16:56, Alan Bateman wrote: On 08/03/2013 15:49, Mike Duigou wrote: Looks fine to me. Do we have an issue open for restoring warnings to the new build? Mike I don't know if there is an issue for that yet but as the new build compiles thousands of classes in a single compilation unit then it means we will need to make significant inroads on the warnings before more can be enabled. The approach with the old build was by area and good progress had been made but with the new build, then it may have to be by warning type as all areas are compiled together. -Alan.
Re: RFR 8009517: Disable fatal compiler warning in the old build
Thank you for trying this Erik. I did think of this workaround myself, but felt if might not be acceptable due to the performance penalty. But this information is great to have. I wonder if we should try to get all alternatives/proposals on the table, then make a decision. I know of two other possibilities. 1) Leave things are they are, and use another tool to investigate warnings. From Jon. 2) Explore supporting Package level SuppressWarnings. Then apply to the relevant packages, and enable -Werror in the build. Others? -Chris. On 11/03/2013 14:20, Erik Joelsson wrote: I tried implementing a PoC for this. Without sjavac, it works, except that the first pass must be run without -Werror and the second with. Since we use -implicit:none, this is fine. With sjavac I had to let it compile the full set of classes first and then run a second time (into a different output directory) with just the warning free set of packages. The overhead on my machine was 5 seconds for the second pass. This solution also works without sjavac, but then the overhead is 37 seconds on my machine. Now the question is, do we want to pursue this or not? /Erik On 2013-03-11 10:30, Erik Joelsson wrote: I have a suggestion for how to at least partly enable -Werror in the new build. The penalty is slightly longer compile time, but the difference should be negligible. We split the big java compilation in jdk in two. The first pass with -Werror and all warnings turned on, the second without. We make a list of packages that are passing -Werror and use as include list for the first and exclude list for the second. As you make more packages warning free, we add them to the list. This solution is not as fine grained as a per package configured set of warning flags, but it's much better than we have today. /Erik On 2013-03-08 16:56, Alan Bateman wrote: On 08/03/2013 15:49, Mike Duigou wrote: Looks fine to me. Do we have an issue open for restoring warnings to the new build? Mike I don't know if there is an issue for that yet but as the new build compiles thousands of classes in a single compilation unit then it means we will need to make significant inroads on the warnings before more can be enabled. The approach with the old build was by area and good progress had been made but with the new build, then it may have to be by warning type as all areas are compiled together. -Alan.
Re: RFR 8009517: Disable fatal compiler warning in the old build
I agree about warning creeping problems. This is a temporary solution, we should soon be fixing the underlying hashcode/equals problems...but... Your temporary solution, -overrides, is just that. It will enable the old build to complete today, but it could fail at any point in the future, as the code changes. For example, java.net is currently warning free, in the old it compiles with fatal warnings enabled. Lets say, in a moment of madness, I add a dependency from java.net.Socket to say java.awt.RenderingHints.Key ( or any class that produces warnings when compiled. I run the new build, all is fine. Push the changes. Now someone else sync's up, but need to build using the old build. If the new dependent class is not already compiled before java.net.Socket gets compiled, it will be compiled implicitly. It's warnings will cause the compile to fail, and the old build will fail. Or much simpler, anyone could write sloppy code with warnings, the new build will suppress them, and they won't notice. Push this code, and the old build will fail if is explicitly, or implicitly, compiles this code with -Werror enabled. We spent a lot of time cleaning up many directories, seems a shame to start allowing non-fatal warnings to come back into previously clean code because people aren't taking the time to fix new warnings as they are introduced. I personally spent several weeks over the past number of years fixing warnings and reviewing warning cleanup webrevs from others. I took much pride in keeping certain areas warnings free. It is with great regret that I propose to disable fatal warnings in the old build, but I felt this the best/safest option. I heard much annoyance and frustration from others about hitting seemingly random errors with the old build recently. This is the only sure way to avoid that. The new builds will still warn, but the old builds will still fail for all but these override problems. Yes, you lose the warnings in the old, but seems better than completely shutting off erroring. I'm ok with that, if others are. To clarify, I think you are suggesting that we keep the old build as it, with -overrides, and use it periodically as a way of tracking new warnings being introduced into areas that were warning free. That is, if the old build fails because of a fatal warning, so be it. File a bug and fix the source code. Then the old build will work again. This means that at any point in time the old build cannot be guaranteed to be buildable. Everyone seems to agree, a solution needs to be found to allow us to keep certain areas warning free. This issue is too important, and too much time was spent, to allow it to regress to the state it was in a few years ago. -Chris. (Ideally it would be nice to warn but not fail on just this one lint option, but don't see how that's possible.) Brad -Chris. Mike On Mar 8 2013, at 05:24 , Chris Hegarty wrote: Since the new build does not enable -Werror when compiling any java code, and disables quite a few lint options, new changes my inadvertently introduce warnings without even realizing. This can cause problems when building with the old build as many areas do compile with -Werror set. Since the old build is on life support, probably best to just completely disable -Werror, so anyone still needing to use it can. diff -r 48b7295f02f8 make/common/shared/Defs-java.gmk --- a/make/common/shared/Defs-java.gmk Thu Mar 07 10:07:13 2013 + +++ b/make/common/shared/Defs-java.gmk Thu Mar 07 11:10:37 2013 + @@ -122,9 +122,10 @@ ifeq ($(JAVAC_MAX_WARNINGS), true) ifeq ($(JAVAC_MAX_WARNINGS), true) JAVAC_LINT_OPTIONS += -Xlint:all endif -ifeq ($(JAVAC_WARNINGS_FATAL), true) - JAVACFLAGS += -Werror -endif +# Disable fatal warnings, 8009517 +#ifeq ($(JAVAC_WARNINGS_FATAL), true) +# JAVACFLAGS += -Werror +#endif # TODO: Workaround for CR 7063027. Remove -path eventually. JAVAC_LINT_OPTIONS += -Xlint:-path -Chris.
Re: RFR 8009517: Disable fatal compiler warning in the old build
On 03/09/2013 12:11 AM, Chris Hegarty wrote: Everyone seems to agree, a solution needs to be found to allow us to keep certain areas warning free. This issue is too important, and too much time was spent, to allow it to regress to the state it was in a few years ago. It is true that selective use of -Werror does not play well with the new build and the desire to compile as much as possible at any time, possibly using sjavac. I have previously reported on a utility I've written in the past to analyze java code and report on the different types of warnings found in different packages. I'm sure we could take something like that utility and improve it to the point where it can give errors when entries in the matrix which should be empty are not. This utility could be run as a test meaning it needn't break the build but the info is available for those that want to run it. -- Jon
Re: RFR 8009517: Disable fatal compiler warning in the old build
On 9 Mar 2013, at 19:01, Jonathan Gibbons jonathan.gibb...@oracle.com wrote: On 03/09/2013 12:11 AM, Chris Hegarty wrote: Everyone seems to agree, a solution needs to be found to allow us to keep certain areas warning free. This issue is too important, and too much time was spent, to allow it to regress to the state it was in a few years ago. It is true that selective use of -Werror does not play well with the new build and the desire to compile as much as possible at any time, possibly using sjavac. I have previously reported on a utility I've written in the past to analyze java code and report on the different types of warnings found in different packages. I'm sure we could take something like that utility and improve it to the point where it can give errors when entries in the matrix which should be empty are not. This utility could be run as a test meaning it needn't break the build but the info is available for those that want to run it. Sounds interesting Jon, and certainly worth exploring. Another idea, is to enable all lint options and -Werror ;-) If the SuppressWarnings annotation was allowable on the PACKAGE Target :-^ It may be reasonable to add to the packages where applicable. -Chris. -- Jon
Re: RFR 8009517: Disable fatal compiler warning in the old build
Looks fine to me. Do we have an issue open for restoring warnings to the new build? Mike On Mar 8 2013, at 05:24 , Chris Hegarty wrote: Since the new build does not enable -Werror when compiling any java code, and disables quite a few lint options, new changes my inadvertently introduce warnings without even realizing. This can cause problems when building with the old build as many areas do compile with -Werror set. Since the old build is on life support, probably best to just completely disable -Werror, so anyone still needing to use it can. diff -r 48b7295f02f8 make/common/shared/Defs-java.gmk --- a/make/common/shared/Defs-java.gmk Thu Mar 07 10:07:13 2013 + +++ b/make/common/shared/Defs-java.gmk Thu Mar 07 11:10:37 2013 + @@ -122,9 +122,10 @@ ifeq ($(JAVAC_MAX_WARNINGS), true) ifeq ($(JAVAC_MAX_WARNINGS), true) JAVAC_LINT_OPTIONS += -Xlint:all endif -ifeq ($(JAVAC_WARNINGS_FATAL), true) - JAVACFLAGS += -Werror -endif +# Disable fatal warnings, 8009517 +#ifeq ($(JAVAC_WARNINGS_FATAL), true) +# JAVACFLAGS += -Werror +#endif # TODO: Workaround for CR 7063027. Remove -path eventually. JAVAC_LINT_OPTIONS += -Xlint:-path -Chris.
Re: RFR 8009517: Disable fatal compiler warning in the old build
On 03/08/2013 08:40 AM, Jonathan Gibbons wrote: On 03/08/2013 08:09 AM, Mike Duigou wrote: On Mar 8 2013, at 07:56 , Alan Bateman wrote: On 08/03/2013 15:49, Mike Duigou wrote: Looks fine to me. Do we have an issue open for restoring warnings to the new build? Mike I don't know if there is an issue for that yet but as the new build compiles thousands of classes in a single compilation unit then it means we will need to make significant inroads on the warnings before more can be enabled. The approach with the old build was by area and good progress had been made but with the new build, then it may have to be by warning type as all areas are compiled together. Understood. Perhaps we can at least use JDK_FILTER incrementally. Do we have a way to override the warnings used by the makefile? Any thoughts towards perhaps disabling -Werror but enabling all of the warnings? Mike A different and maybe more effective way of tracking this (for now) would be to generate a separate report on a regular basis that details the number of different types of warnings in each package. The technology to do that is easy; the hard part is getting people to monitor the results and fixup the issues. That being said, it should be relatively uncontroversial to totally eliminate some types of warnings, like cast warnings, and enable those warnings in the build with -Werror. -- Jon Kurchi and I are working on this as our side project. We have modified your java program to analyse the build logs generated by the new build. I am thinking to generate some trend reports after running the build regularly and collecting some log data. -Dan
Re: RFR 8009517: Disable fatal compiler warning in the old build
I responded in another thread (wasn't aware of this one, sorry), there is an alternate to completely disabling -Werror. On 3/8/2013 7:58 AM, Chris Hegarty wrote: On 08/03/2013 15:49, Mike Duigou wrote: Looks fine to me. Thanks Mike. Do we have an issue open for restoring warnings to the new build? Not yet, that I am aware of. We really need the ability to set lint options per package/subpackage. That would be nice. I agree about warning creeping problems. This is a temporary solution, we should soon be fixing the underlying hashcode/equals problems...but... 1. javac tightened hashcode/equals checks 2. new: -Werror is off in the new builds. (i.e. not failing on any lint warnings) 3. old: -Werror is on for the old builds (i.e. is failing for any lint warnings) The proposal is to turn off all errors in the old builds (remove -Werror), essentially making 3 like the 2. Warn but not fatal. We spent a lot of time cleaning up many directories, seems a shame to start allowing non-fatal warnings to come back into previously clean code because people aren't taking the time to fix new warnings as they are introduced. My suggestion was to turn off just the one warning type in 3 *in the old code only* so we can at least continue to build the old without completely disabling -Werror. The new builds will still warn, but the old builds will still fail for all but these override problems. Yes, you lose the warnings in the old, but seems better than completely shutting off erroring. (Ideally it would be nice to warn but not fail on just this one lint option, but don't see how that's possible.) Brad -Chris. Mike On Mar 8 2013, at 05:24 , Chris Hegarty wrote: Since the new build does not enable -Werror when compiling any java code, and disables quite a few lint options, new changes my inadvertently introduce warnings without even realizing. This can cause problems when building with the old build as many areas do compile with -Werror set. Since the old build is on life support, probably best to just completely disable -Werror, so anyone still needing to use it can. diff -r 48b7295f02f8 make/common/shared/Defs-java.gmk --- a/make/common/shared/Defs-java.gmk Thu Mar 07 10:07:13 2013 + +++ b/make/common/shared/Defs-java.gmk Thu Mar 07 11:10:37 2013 + @@ -122,9 +122,10 @@ ifeq ($(JAVAC_MAX_WARNINGS), true) ifeq ($(JAVAC_MAX_WARNINGS), true) JAVAC_LINT_OPTIONS += -Xlint:all endif -ifeq ($(JAVAC_WARNINGS_FATAL), true) - JAVACFLAGS += -Werror -endif +# Disable fatal warnings, 8009517 +#ifeq ($(JAVAC_WARNINGS_FATAL), true) +# JAVACFLAGS += -Werror +#endif # TODO: Workaround for CR 7063027. Remove -path eventually. JAVAC_LINT_OPTIONS += -Xlint:-path -Chris.
