Push a real iseq in rb_vm_push_frame_fname() Previously, vm_make_env_each() (used during proc creation and for the debug inspector C API) picked up the non-GC-allocated iseq that rb_vm_push_frame_fname() creates, which led to a SEGV when the GC tried to mark the non GC object. Put a real iseq imemo instead. Speed should be about the same since the old code also did a imemo allocation and a malloc allocation. Real iseq allows ironing out the special-casing of dummy frames in rb_execution_context_mark() and rb_execution_context_update(). A check is added to RubyVM::ISeq#eval, though, to stop attempts to run dummy iseqs. [Bug #21180] Co-authored-by: Aaron Patterson <tenderlove@ruby-lang.org>
[Bug #18257] Register the class path of FrozenCore to mark ICLASS does not have the path usually, so it needs to be registered separately.
Https://learn.microsoft.com/en-us/cpp/build/reference/zc-inline-remove-unreferenced-comdat?view=msvc-140 > If `/Zc:inline` is specified, the compiler enforces the C++11 > requirement that all functions declared inline must have a definition > available in the same translation unit if they're used.
[Bug #19592] Fix ext/Setup support After [1], using ext/Setup to link some, but not all extensions failed during linking. I did not know about this option, and had assumed that only `--with-static-linked-ext` builds can include statically linked extensions. Include the support code for statically linked extensions in all configurations like before [1]. Initialize the table lazily to minimize footprint on builds that have no statically linked extensions. [1]: 790cf4b6d0475614afb127b416e87cfa39044d67 "Fix autoload status of statically linked extensions" --- load.c | 28 +++++++++++++++++----------- vm.c | 3 --- vm_core.h | 6 ------ 3 files changed, 17 insertions(+), 20 deletions(-)
Speed up rebuilding the loaded feature index Rebuilding the loaded feature index slowed down with the bug fix for #17885 in 79a4484a072e9769b603e7b4fbdb15b1d7eccb15. The slowdown was extreme if realpath emulation was used, but even when not emulated, it could be about 10x slower. This adds loaded_features_realpath_map to rb_vm_struct. This is a hidden hash mapping loaded feature paths to realpaths. When rebuilding the loaded feature index, look at this hash to get cached realpath values, and skip calling rb_check_realpath if a cached value is found. Fixes [Bug #19246] --- load.c | 27 +++++++++++++++++++++++---- vm.c | 2 ++ vm_core.h | 1 + 3 files changed, 26 insertions(+), 4 deletions(-)
This reverts commit 788b03d5ba82fd8b35ce1fe2618ce6bacc648333.
* Re-apply "Ruby 3.2 - Speed up rebuilding the loaded feature index and realpath cache (#8023)" * [CI] mingw.yml - remove IBM437 encoding for test-all, use cmd shell for test & test-all * Skip failing test on mingw with readline.so Co-authored-by: nagachika <nagachika@ruby-lang.org> --------- Co-authored-by: nagachika <nagachika@ruby-lang.org>
This reverts commit 8346d1630b8193eef1ec9dd537b16de74afdc2e8. It seems break CI on MinGW. https://github.com/ruby/ruby/actions/runs/5698775895/job/15447455988
* Speed up rebuilding the loaded feature index Rebuilding the loaded feature index slowed down with the bug fix for #17885 in 79a4484a072e9769b603e7b4fbdb15b1d7eccb15. The slowdown was extreme if realpath emulation was used, but even when not emulated, it could be about 10x slower. This adds loaded_features_realpath_map to rb_vm_struct. This is a hidden hash mapping loaded feature paths to realpaths. When rebuilding the loaded feature index, look at this hash to get cached realpath values, and skip calling rb_check_realpath if a cached value is found. Fixes [Bug #19246] * Add a realpath cache to reduce number of syscalls. Number of lstat and stat syscalls for each 'require'd file is doubled, because rb_realpath_internal is called from two places with the same arguments in require_internal; once for checking the realpaths cache, and once in load_iseq_eval when iseq is not found. Introduce rb_realpath_internal_cached function to reuse the realpath_map cache which memoizes rb_realpath_internal function, leading to less syscalls and increased startup performance depending on the cost of the syscalls in a particular environment. --------- Co-authored-by: Jeremy Evans <code@jeremyevans.net>
Fix off-by-one error in rb_vm_each_stack_value Applying the following patch to test/erb/test_erb.rb and running that file will cause Ruby to crash on my machine (macOS 13.1 on M1 Pro): ``` --- a/test/erb/test_erb.rb +++ b/test/erb/test_erb.rb @@ -7,6 +7,12 @@ class TestERB < Test::Unit::TestCase class MyError < RuntimeError ; end + def setup + GC.auto_compact = true + GC.stress = true + GC.verify_compaction_references(expand_heap: true, toward: :empty) + end + ``` It crashes with the following log: ``` /Users/peter/src/ruby/lib/erb/compiler.rb:276: [BUG] Segmentation fault at 0x00000001083a8690 ... -- C level backtrace information ------------------------------------------- ... /Users/peter/src/ruby/build/ruby(rb_vm_each_stack_value+0xa8) [0x104cc3a44] ../vm.c:2737 /Users/peter/src/ruby/build/ruby(rb_vm_each_stack_value+0xa8) [0x104cc3a44] ../vm.c:2737 /Users/peter/src/ruby/build/ruby(check_stack_for_moved+0x2c) [0x104b272a4] ../gc.c:5512 /Users/peter/src/ruby/build/ruby(gc_compact_finish) ../gc.c:5534 /Users/peter/src/ruby/build/ruby(gc_sweep_compact) ../gc.c:8653 /Users/peter/src/ruby/build/ruby(gc_sweep) ../gc.c:6196 /Users/peter/src/ruby/build/ruby(has_sweeping_pages+0x0) [0x104b19c54] ../gc.c:9568 /Users/peter/src/ruby/build/ruby(gc_rest) ../gc.c:9570 ``` This crash happens because it's reading the VALUE at sp. But since sp points to the top of the stack, it's reading the VALUE above the top of the stack, which is causing this segfault. Fixes [Bug #19320] --- vm.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)