Replace alloc-related offset type to usize instead isize

[fepicture]

EDIT1: fix #674
this pr only replaces the isize to usize for offset in allocation, which makes the code more robust.
there is no further test since a negative will be a failure at compile time.

[fepicture]

well I have run the .github/script/ci-test.sh, and encountered errors in compiling pfm-sys(third-party lib), I do not think this is related to mmtk, say the full script does not run.

error log

++ [[ x86_64 == \x\8\6\_\6\4 ]]
++ [[ linux == \l\i\n\u\x ]]
++ cargo test --features perf_counter
   Compiling pfm-sys v0.0.14
   Compiling mmtk v0.18.0 (/mmtk-core)
error: failed to run custom build command for `pfm-sys v0.0.14`



Caused by:
  process didn't exit successfully: `/mmtk-core/target/debug/build/pfm-sys-1aa429483a40e9a0/build-script-build` (exit status: 101)
  --- stdout
  OUT_DIR: "/mmtk-core/target/debug/build/pfm-sys-4c4afeaa387dfb3f/out"
  CARGO_MANIFEST_DIR: "/.cargo/registry/src/github.com-1ecc6299db9ec823/pfm-sys-0.0.14"
  Compiling for 'x86_64' target
  Compiling for 'Linux' system
  make[1]: Entering directory '/mmtk-core/target/debug/build/pfm-sys-4c4afeaa387dfb3f/out/libpfm4/lib'
  clang-15 -fPIC  -g -Wall -Werror -Wextra -Wno-unused-parameter -I. -I/mmtk-core/target/debug/build/pfm-sys-4c4afeaa387dfb3f/out/libpfm4/include -DCONFIG_PFMLIB_DEBUG -DCONFIG_PFMLIB_OS_LINUX  -g -Wall -Werror -Wextra -Wno-unused-parameter -I. -I/mmtk-core/target/debug/build/pfm-sys-4c4afeaa387dfb3f/out/libpfm4/lib/../include -DCONFIG_PFMLIB_DEBUG -DCONFIG_PFMLIB_OS_LINUX -D_REENTRANT -I. -fvisibility=hidden -DCONFIG_PFMLIB_ARCH_X86 -DCONFIG_PFMLIB_ARCH_X86_64 -I. -c pfmlib_common.c
  make[1]: Leaving directory '/mmtk-core/target/debug/build/pfm-sys-4c4afeaa387dfb3f/out/libpfm4/lib'

  --- stderr
  pfmlib_common.c:1780:12: error: variable 'n' set but not used [-Werror,-Wunused-but-set-variable]
          int i, u, n = 0, um;
                    ^
  1 error generated.
  make[1]: *** [/mmtk-core/target/debug/build/pfm-sys-4c4afeaa387dfb3f/out/libpfm4/lib/../rules.mk:30: pfmlib_common.o] Error 1
  make: *** [Makefile:49: all] Error 2
  thread 'main' panicked at 'make exited with status exit status: 2', /.cargo/registry/src/github.com-1ecc6299db9ec823/pfm-sys-0.0.14/build.rs:47:9
  stack backtrace:
     0: rust_begin_unwind
               at /rustc/90743e7298aca107ddaa0c202a4d3604e29bfeb6/library/std/src/panicking.rs:575:5
     1: core::panicking::panic_fmt
               at /rustc/90743e7298aca107ddaa0c202a4d3604e29bfeb6/library/core/src/panicking.rs:65:14
     2: build_script_build::main
               at ./build.rs:47:9
     3: core::ops::function::FnOnce::call_once
               at /rustc/90743e7298aca107ddaa0c202a4d3604e29bfeb6/library/core/src/ops/function.rs:251:5
  note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.
warning: build failed, waiting for other jobs to finish...
$ ./ci-test.sh 

GitHub mirror old code

source forge latest code have solved this one?

EDIT1:

I got it. CC @caizixian
I use clang-15 to run ci-test.sh which need compile pfm package, and encounter error state the unused var.
pfm is a high-level wrapper about libpfm4.
here libpfm4 recently adds a new commit to remove the unused var which can solve this problem. https://sourceforge.net/p/perfmon2/libpfm4/ci/533633adf7d00bbfcb7f2759567869d585bf97e1/

and our cargo used pfm which is from https://github.com/caizixian/pfm, which actually uses a git submodule to control the internal libpfm4 code, simply updating the submodule commit may solve this? may @caizixian help to update? the submodule i cannot access..caizixian/pfm@7223ed8

[caizixian]

@fepicture fixed. Please refer to the latest commit.

[fepicture]

May you have time to take a review? @qinsoon
the failure ci is public API change， this is normal right？

Btw, finally, I used gcc11 to build pfm(clang15 has other pitches, and the upstream has not changed to fix)

[qinsoon]

May you have time to take a review? @qinsoon the failure ci is public API change， this is normal right？

The PR looks good. The failure in the API check is expected, as the PR changes the API. I will submit a few PRs to update our bindings for this change. Thanks again for the PR.

[k-sareen]

Could you change this file as well? The offset is a ssize_t here: https://github.com/mmtk/mmtk-core/blob/c5a20c059674c4eee2c1d005180b1f12a99088e1/vmbindings/dummyvm/api/mmtk.h

[k-sareen]

Apparently there's another header file here: https://github.com/mmtk/mmtk-core/blob/c5a20c059674c4eee2c1d005180b1f12a99088e1/docs/header/mmtk.h

It should be updated as well.

[fepicture]

Apparently there's another header file here: https://github.com/mmtk/mmtk-core/blob/c5a20c059674c4eee2c1d005180b1f12a99088e1/docs/header/mmtk.h

It should be updated as well.

Thank you for your review and for pointing out the omission!
I have made the necessary changes to the file as suggested.
Again, Thank you for your patience and guidance.

[qinsoon]

binding-refs
OPENJDK_BINDING_REPO=qinsoon/mmtk-openjdk
OPENJDK_BINDING_REF=update-mmtk-core-838
JIKESRVM_BINDING_REPO=qinsoon/mmtk-jikesrvm
JIKESRVM_BINDING_REF=update-mmtk-core-838
JULIA_BINDING_REPO=qinsoon/mmtk-julia
JULIA_BINDING_REF=update-mmtk-core-838

[fepicture]

binding-refs OPENJDK_BINDING_REPO=qinsoon/mmtk-openjdk OPENJDK_BINDING_REF=update-mmtk-core-838 JIKESRVM_BINDING_REPO=qinsoon/mmtk-jikesrvm JIKESRVM_BINDING_REF=update-mmtk-core-838 JULIA_BINDING_REPO=qinsoon/mmtk-julia JULIA_BINDING_REF=update-mmtk-core-838

I am really sorry to change the commit info by rebase to solve the out of date,
may those dependencies need extra updates? https://github.com/mmtk/mmtk-openjdk/pull/221/files#diff-99ac2fb415f1a54ae4d65e0487c56226749e91ad44f9fffcddccefa38de8e1a0R33
so If there out-of-date master branch, can I solve this by myself(outer-contributer)?
say we only use merge and do not use rebase to avoid the original commit change?

[qinsoon]

binding-refs OPENJDK_BINDING_REPO=qinsoon/mmtk-openjdk OPENJDK_BINDING_REF=update-mmtk-core-838 JIKESRVM_BINDING_REPO=qinsoon/mmtk-jikesrvm JIKESRVM_BINDING_REF=update-mmtk-core-838 JULIA_BINDING_REPO=qinsoon/mmtk-julia JULIA_BINDING_REF=update-mmtk-core-838

I am really sorry to change the commit info by rebase to solve the out of date, may those dependencies need extra updates? https://github.com/mmtk/mmtk-openjdk/pull/221/files#diff-99ac2fb415f1a54ae4d65e0487c56226749e91ad44f9fffcddccefa38de8e1a0R33 so If there out-of-date master branch, can I solve this by myself(outer-contributer)? say we only use merge and do not use rebase to avoid the original commit change?

I will update the binding PRs once this mmtk-core PR is merged. Don't worry about it.

[qinsoon]

LGTM. Thanks for the PR.