Add option to use pool implemenation which relies on memalign #4357
Conversation
ponylang-main
added
the
discuss during sync
|
Is there a particular memory issue etc that you are planning on looking into when this goes in? |
src/libponyc/pass/names.c
Outdated
| ast = *astp; | ||
|
|
There was a problem hiding this comment.
is there a known bug from this?
There was a problem hiding this comment.
not that i'm aware of.. just something that the memory sanitizer flagged (iirc)..
There was a problem hiding this comment.
#4360 created for this
src/libponyc/ast/lexer.c
Outdated
| @@ -607,7 +607,7 @@ static void normalise_string(lexer_t* lexer) | |||
|
|
|||
| // Trim trailing empty line | |||
| size_t trim = 0; | |||
| for (size_t i = (lexer->buflen - 1); i>0; i--) | |||
| for (ssize_t i = (lexer->buflen - 1); i>0; i--) | |||
There was a problem hiding this comment.
is there a known bug from this?
There was a problem hiding this comment.
not that i'm aware of.. just something that the undefined behavior sanitizer flagged (iirc)..
There was a problem hiding this comment.
#4359 created for this
|
Are the two small bugfixes required for this PR? They appear "unrelated-ish" and should probably going as 1 or 2 separate PRs. |
|
Is "fix to make sure the test runner gets rebuilt whenever ponyc gets rebuilt" a 3rd separate bug fix that should also be it's own PR? |
not specifically.. i added this because i was running into some weird crashes on something i'm trying out and i was trying to rule out potential causes and wanted to rely on the memory sanitizer as sanity check.. |
Not required.. i'll bump them into their own PRs.. |
Yep, it can be it's own PR.. i'll slim this down to only what's needed for the new functionality and break everything else out into separate PRs.. |
Awesome |
|
@dipinhora can you rebase this against main? |
done |
|
@jemc any objections to this PR? |
This commit add an option to build pony with the pool implementation based on posix_memalign instead of the default pool implementation. This, combined with address/undefined behavior/etc sanitizers, allows for easier debugging of memory issues within the compiler and runtime. This commit also includes two small bugfixes in the compiler that were discovered as part of this: * lexer.c fix in `normalise_string` to prevent issues when `lexer->buflen` == 0 * names.c fix in `names_typeparam` to prevent access to an AST node that was just replaced and a fix to the `Makefile` for running tests with `use=` flags and a fix to make sure the test runner gets rebuilt whenever ponyc gets rebuilt
|
rebased onto |
ponylang-main
removed
the
discuss during sync
|
@SeanTAllen i made a mistake and my last rebase to this PR undid all the changes that were split off into their own PRs and now
how do you want to proceed? should i create 5 new PRs to re-apply them or should i create a single one with all the changes combined? |
ponylang-main
added
the
discuss during sync
|
I'll roll this one back later today then redo just this one. |
|
@SeanTAllen #4367 has been created to do the |
This commit add an option to build pony with the pool implementation based on posix_memalign instead of the default pool implementation. This, combined with address/undefined behavior/etc sanitizers, allows for easier debugging of memory issues within the compiler and runtime. This commit also includes two small bugfixes in the compiler that were discovered as part of this:
normalise_stringto prevent issues whenlexer->buflen== 0names_typeparamto prevent access to an AST node that was just replacedand a fix to the
Makefilefor running tests withuse=flags and a fix to make sure the test runner gets rebuilt whenever ponyc gets rebuilt