-
Notifications
You must be signed in to change notification settings - Fork 48
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Delete classpath argument files upon shutdown. Fixes #240 #242
Conversation
org.eclipse.jdt.launching/launching/org/eclipse/jdt/internal/launching/ClasspathShortener.java
Show resolved
Hide resolved
@@ -313,6 +317,25 @@ private void shortenPathUsingClasspathArgumentFile(int modulePathArgumentIndex, | |||
addProcessTempFile(file); | |||
} | |||
|
|||
private void createOrUpdateGitIgnore() { | |||
File gitIgnore = new File(processTempFilesDir, ".gitignore"); //$NON-NLS-1$ |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Instead of touching gitignore, one can contribute something like:
<extension
point="org.eclipse.team.core.ignore">
<ignore
enabled="true"
pattern=".temp-*path-arg-*.txt"/>
</extension>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Unfortunately neither ".temp-*path-arg-*.txt
" nor "**/.temp-*path-arg-*.txt
" actually hides these files in the Package Explorer; I guess I have to investigate further.
So this is just to actually have these files ignored by version control, right?
The original intention was to hide them, so which solution is better?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please also note that there are multiple ways to call org.eclipse.jdt.internal.launching.JavaLaunchingUtils.createFileForArgument(String, File, String, String)
which may result in different patterns. As far as i remember one usecase is to have a long classpath while another usecase is more general to have over-long parameters at all.
@iloveeclipse would it be feasible to use a dedicated temp path below
workspace\.metadata\.plugins\org.eclipse.jdt.launching
?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
would it be feasible to use a dedicated temp path below
workspace.metadata.plugins\org.eclipse.jdt.launching?
Technically we can create these temp files there, yes. I wonder why that wasn't done and if there is a catch (like having them in the base directory etc).
@diti0023 : please don't use merge to get latest state from master, please use rebase. |
@iloveeclipse Thanks -- obviously I'm still a newbie ... |
ea40335
to
6be5268
Compare
Current solution: |
Note: we are in 4.28 RC phase now and I personally would prefer that we would postpone this PR to 4.29 M1. |
/rebase |
It is not temporary issue.
If this is the first change in 4.29 on org.eclipse.jdt.launching project, the version should be bumped by +100. |
org.eclipse.jdt.launching/launching/org/eclipse/jdt/internal/launching/ClasspathShortener.java
Outdated
Show resolved
Hide resolved
* Remove use of OPTION_JdtDebugCompileMode with OPTION_IgnoreUnnamedModuleForSplitPackage This change replaces use of OPTION_JdtDebugCompileMode with OPTION_IgnoreUnnamedModuleForSplitPackage, since the latter does what we need better and doesn't cause the side effects that the first option causes. Goal is to later on remove OPTION_JdtDebugCompileMode in JDT core. Fixes: #260 Signed-off-by: Simeon Andreev <simeon.danailov.andreev@gmail.com> * Detect JVM installs at startup (#231) * Detect JVM installs at startup Fixes #230 * Delete classpath argument files upon shutdown. Fixes #240 (#242) Store classpath argument files in bundle state location, not in workspace - Avoids that workspace is contaminated with temporary files. - Delete classpath argument files upon shutdown. - Version bump org.eclipse.jdt.launching. See https://github.com/eclipse-platform/eclipse.platform.debug/issues/84 Fixes #240 * Avoid reusing of not thread-safe SimpleDateFormat SimpleDateFormat.format(new Date()); returns same string as immutable and thread-safe DateTimeFormatter.format(new Date().toInstant()) * Avoid potential Threadlocal.set(null) memory leak (#249) Threadlocal.set(null) keeps a reference to ThreadLocal.this see https://rules.sonarsource.com/java/RSPEC-5164 Co-authored-by: Jörg Kubitz <jkubitz-eclipse@gmx.de> * Change how JrtPackageFragmentRoots are printed in Source tab (#266) - currently all JrtPackageFragmentRoots are printed using the path which is the same for all modules extracted from a jrt-fs.jar - change this to print the module name followed by the full jrt-fs.jar path - fixes #633 * JavaSnippetEditor: fix missing synchronize (#259) * JavaSnippetEditor: fix missing synchronize * JavaSnippetEditor: Adding "this." for non static field access --------- Co-authored-by: Jörg Kubitz <jkubitz-eclipse@gmx.de> * LaunchingPlugin: use try-with-resource (#252) to close streams in case of Exceptions. Co-authored-by: Jörg Kubitz <jkubitz-eclipse@gmx.de> * Add missing "static" modifier for constants (#256) To reduce memory per instance Co-authored-by: Jörg Kubitz <jkubitz-eclipse@gmx.de> * [trivial] Fixed "Redundant specification of type arguments" warnings Reported by official SDK build since 4.29 for whatever reason See for example https://download.eclipse.org/eclipse/downloads/drops4/I20230621-1800/compilelogs/plugins/org.eclipse.jdt.debug.jdi.tests_1.1.0.v20230328-1614/@dot.html#OTHER_WARNINGS * Records with inner records breaks breakpoint toggling fixes #270 Boundary for Breakpoint should be checked * Touch bundles affected by the new ecj version See eclipse-platform/eclipse.platform.releng.aggregator#1184 * Fix html links in VMInstall(Type) Extension-Point documentation * Re-throw underlying exceptions in OpenFromClipboardTests #27 (#279) Replace call to Display::syncExec with call to Display::syncCall in the private method getJavaElementMatches. snycCall does not hide exceptions, it re-throws them to the calling thread (which helps debugging). Don't catch exceptions in the helper class "Accessor", re-throw them so the calling tests can show a more meaningful stack-trace when they fail. * Use typed for-each loop Avoiding type-casts makes code easier to read. * Avoid wrapping primitives for toString() see https://rules.sonarsource.com/java/RSPEC-2131 * ReferenceTypeImpl: compare Array content instead of instance comparison. https://stackoverflow.com/questions/8777257/equals-vs-arrays-equals-in-java * VMInstallTests: fix random ConcurrentModificationException #202 by using threadsafe datastructures. * Set console stream encoding for java >= 19 - eclipse.platform.debug#124 https://github.com/eclipse-platform/eclipse.platform.debug/issues/124 * FileHashing: avoid getCanonicalFile() getCanonicalFile() is slow on windows / JDK17 - can costs even more then calculating the SHA1 of the file content, which it is supposed to avoid. * Instead use normalized path. It does not matter that multiple files could be the same due to symbolic links, it will just recalculate their SHA1 for all of them - which is still faster. * Also read all file attributes at once. * Refactor XML parsing * Fix path decoding of URL to File containing space (fixes #64) A space in an URL is encoded as "%20" and has to be decoded to " ". Otherwise a debugged JAR rooted in a path containing space is not opening class file but java file editor. * Add MacOS support for detecting VM installs at start up (#276) - fixes #265 * require org.eclipse.core.runtime 3.29.0 --------- Signed-off-by: Simeon Andreev <simeon.danailov.andreev@gmail.com> Co-authored-by: Simeon Andreev <simeon.danailov.andreev@gmail.com> Co-authored-by: Mickael Istria <mistria@redhat.com> Co-authored-by: Diethard Ohrt <130975899+diti0023@users.noreply.github.com> Co-authored-by: Jörg Kubitz <jkubitz-eclipse@gmx.de> Co-authored-by: Jörg Kubitz <51790620+jukzi@users.noreply.github.com> Co-authored-by: Jeff Johnston <jjohnstn@redhat.com> Co-authored-by: Andrey Loskutov <loskutov@gmx.de> Co-authored-by: Hannes Wellmann <wellmann.hannes1@gmx.net> Co-authored-by: Federico Jeanne <2205684+fedejeanne@users.noreply.github.com>
* Remove use of OPTION_JdtDebugCompileMode with OPTION_IgnoreUnnamedModuleForSplitPackage This change replaces use of OPTION_JdtDebugCompileMode with OPTION_IgnoreUnnamedModuleForSplitPackage, since the latter does what we need better and doesn't cause the side effects that the first option causes. Goal is to later on remove OPTION_JdtDebugCompileMode in JDT core. Fixes: #260 Signed-off-by: Simeon Andreev <simeon.danailov.andreev@gmail.com> * Detect JVM installs at startup (#231) * Detect JVM installs at startup Fixes #230 * Delete classpath argument files upon shutdown. Fixes #240 (#242) Store classpath argument files in bundle state location, not in workspace - Avoids that workspace is contaminated with temporary files. - Delete classpath argument files upon shutdown. - Version bump org.eclipse.jdt.launching. See https://github.com/eclipse-platform/eclipse.platform.debug/issues/84 Fixes #240 * Avoid reusing of not thread-safe SimpleDateFormat SimpleDateFormat.format(new Date()); returns same string as immutable and thread-safe DateTimeFormatter.format(new Date().toInstant()) * Avoid potential Threadlocal.set(null) memory leak (#249) Threadlocal.set(null) keeps a reference to ThreadLocal.this see https://rules.sonarsource.com/java/RSPEC-5164 Co-authored-by: Jörg Kubitz <jkubitz-eclipse@gmx.de> * Change how JrtPackageFragmentRoots are printed in Source tab (#266) - currently all JrtPackageFragmentRoots are printed using the path which is the same for all modules extracted from a jrt-fs.jar - change this to print the module name followed by the full jrt-fs.jar path - fixes #633 * JavaSnippetEditor: fix missing synchronize (#259) * JavaSnippetEditor: fix missing synchronize * JavaSnippetEditor: Adding "this." for non static field access --------- Co-authored-by: Jörg Kubitz <jkubitz-eclipse@gmx.de> * LaunchingPlugin: use try-with-resource (#252) to close streams in case of Exceptions. Co-authored-by: Jörg Kubitz <jkubitz-eclipse@gmx.de> * Add missing "static" modifier for constants (#256) To reduce memory per instance Co-authored-by: Jörg Kubitz <jkubitz-eclipse@gmx.de> * [trivial] Fixed "Redundant specification of type arguments" warnings Reported by official SDK build since 4.29 for whatever reason See for example https://download.eclipse.org/eclipse/downloads/drops4/I20230621-1800/compilelogs/plugins/org.eclipse.jdt.debug.jdi.tests_1.1.0.v20230328-1614/@dot.html#OTHER_WARNINGS * Records with inner records breaks breakpoint toggling fixes #270 Boundary for Breakpoint should be checked * Touch bundles affected by the new ecj version See eclipse-platform/eclipse.platform.releng.aggregator#1184 * Fix html links in VMInstall(Type) Extension-Point documentation * Re-throw underlying exceptions in OpenFromClipboardTests #27 (#279) Replace call to Display::syncExec with call to Display::syncCall in the private method getJavaElementMatches. snycCall does not hide exceptions, it re-throws them to the calling thread (which helps debugging). Don't catch exceptions in the helper class "Accessor", re-throw them so the calling tests can show a more meaningful stack-trace when they fail. * Use typed for-each loop Avoiding type-casts makes code easier to read. * Avoid wrapping primitives for toString() see https://rules.sonarsource.com/java/RSPEC-2131 * ReferenceTypeImpl: compare Array content instead of instance comparison. https://stackoverflow.com/questions/8777257/equals-vs-arrays-equals-in-java * VMInstallTests: fix random ConcurrentModificationException #202 by using threadsafe datastructures. * Set console stream encoding for java >= 19 - eclipse.platform.debug#124 https://github.com/eclipse-platform/eclipse.platform.debug/issues/124 * FileHashing: avoid getCanonicalFile() getCanonicalFile() is slow on windows / JDK17 - can costs even more then calculating the SHA1 of the file content, which it is supposed to avoid. * Instead use normalized path. It does not matter that multiple files could be the same due to symbolic links, it will just recalculate their SHA1 for all of them - which is still faster. * Also read all file attributes at once. * Refactor XML parsing * Fix path decoding of URL to File containing space (fixes #64) A space in an URL is encoded as "%20" and has to be decoded to " ". Otherwise a debugged JAR rooted in a path containing space is not opening class file but java file editor. * Add MacOS support for detecting VM installs at start up (#276) - fixes #265 * require org.eclipse.core.runtime 3.29.0 --------- Signed-off-by: Simeon Andreev <simeon.danailov.andreev@gmail.com> Co-authored-by: Simeon Andreev <simeon.danailov.andreev@gmail.com> Co-authored-by: Mickael Istria <mistria@redhat.com> Co-authored-by: Diethard Ohrt <130975899+diti0023@users.noreply.github.com> Co-authored-by: Jörg Kubitz <jkubitz-eclipse@gmx.de> Co-authored-by: Jörg Kubitz <51790620+jukzi@users.noreply.github.com> Co-authored-by: Jeff Johnston <jjohnstn@redhat.com> Co-authored-by: Andrey Loskutov <loskutov@gmx.de> Co-authored-by: Hannes Wellmann <wellmann.hannes1@gmx.net> Co-authored-by: Federico Jeanne <2205684+fedejeanne@users.noreply.github.com>
View also original issue for more information.
What it does
In the dialog Run Configurations, click on "Show Comandline". If there are many or long classpaths, they are collected in a file in order not to exceed the maximum command length (on a Linux system check with
getconf ARG_MAX
). This is also done for actually launching the run, but in that case these files are removed on terminating the launch. When just showing the command line these files remain, in order to have them accessible later (i.e. after closing the dialog). In this case, i.e. without any process, the GUI has no access to these files. This change marks the files as to be deleted on shutdown of the JVM, and handles them via a .gitignore in order to prevent accidentally committing them.How to test
Create a project with many JAR files and long paths to them. When the command length of the OS is about to be exceeded, the resulting command changes from listing all files as parameters (-classpath) to a single entry like
@/home/.../.temp-Main--module-path-arg-1681985065801.txt
The files resulting from just showing the commandline are deleted when the JVM is terminated.
Author checklist