long_subtype_new underallocates for zero #101037
Comments
mdickinson
added
type-bug
mdickinson
added a commit
to mdickinson/cpython
that referenced
this issue
Jan 14, 2023
corona10
pushed a commit
to mdickinson/cpython
that referenced
this issue
Jan 19, 2023
mdickinson
added a commit
that referenced
this issue
Jan 21, 2023
…ypes (#101038) This PR fixes object allocation in long_subtype_new to ensure that there's at least one digit in all cases, and makes sure that the value of that digit is copied over from the source long. Needs backport to 3.11, but not any further: the change to require at least one digit was only introduced for Python 3.11. Fixes #101037.
miss-islington
pushed a commit
to miss-islington/cpython
that referenced
this issue
Jan 21, 2023
…t subtypes (pythonGH-101038) This PR fixes object allocation in long_subtype_new to ensure that there's at least one digit in all cases, and makes sure that the value of that digit is copied over from the source long. Needs backport to 3.11, but not any further: the change to require at least one digit was only introduced for Python 3.11. Fixes pythonGH-101037. (cherry picked from commit 401fdf9) Co-authored-by: Mark Dickinson <dickinsm@gmail.com>
mdickinson
added a commit
that referenced
this issue
Jan 21, 2023
…nt subtypes (GH-101038) (#101219) gh-101037: Fix potential memory underallocation for zeros of int subtypes (GH-101038) This PR fixes object allocation in long_subtype_new to ensure that there's at least one digit in all cases, and makes sure that the value of that digit is copied over from the source long. Needs backport to 3.11, but not any further: the change to require at least one digit was only introduced for Python 3.11. Fixes GH-101037. (cherry picked from commit 401fdf9) Co-authored-by: Mark Dickinson <dickinsm@gmail.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
When creating
intinstances, we always allocate space for and initialise at least one digit, since themedium_valuefunction that's used for fast paths assumes the existence of that digit. This change was introduced for Python 3.11 in PR #27832.However, the corresponding change in
long_subtype_newwas missed; this results in us asking for zero digits in the case of creating an instance of a subclass ofintwith value zero - e.g.,That's then a problem if
myintis used in a context wheremedium_valuemight be called on it, for example as inint(myint).We've got away with this so far because of a compensating overallocation issue: see #81381; this bug blocked an attempt to address #81381 in #100855. But I believe this should still be fixed.
I'll make a PR shortly.
Linked PRs
The text was updated successfully, but these errors were encountered: