git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH v2] grep: fall back to interpreter if JIT memory allocation fails

From
Junio C Hamano <gitster@pobox.com>
Date
Jan 27, 2023, 17:39 UTC
Message-ID
<xmqq1qnfancf.fsf@gitster.g>
In-Reply-To
<xmqqbkmk9bsn.fsf@gitster.g>
Junio C Hamano <gitster@pobox.com> writes:
Show 10 quoted lines
> Yes, the "instead of failing hard, fall back" makes sense.  Just
> that I do not see why the runtime test is a good thing to have.  In
> short, we are not in the business of catching bugs in pcre2_jit
> implementations, so if they say they cannot compile the pattern (I
> would even say I doubt the point of checking the return code to
> ensure it is NOMEMORY), it would be fine to let the interpreter
> codepath to inspect the pattern and diagnose problems with it, or
> take the slow match without JIT.
>
> What am I missing?

Note that I've seen and recently re-read the discussion that leads to https://lore.kernel.org/git/f680b274-fa85-6624-096a-7753a2671c15@grsecurity.net/

I suspect that this auto-probe is related to solving "the user thinks JIT is in use but because of failing JIT the user's pattern is getting horrible performance" somehow. But I do not think a hard failure is a good approach to help users in such a situation.

After such a failure, the user can prefix "(*NO_JIT)" to the pattern and retry, or give up the operation altogether and not get a useful result, but wouldn't it be far more helpful to just fallback as if (*NO_JIT) was on from the beginning?

Also I notice that p->pcre2_jit_on is per "struct grep_pat", so it is not like "once we see a pathological pattern, we turn off JIT completely for other patterns", right? That is, if you have

    git grep -P -e "$A" -e "$B"

and we fail to compile "$A" (for whatever reason), we could still (attempt to) compile "$B". Perhaps $A was too complex or was incompatible with JIT combined with other options, but $B may be easy enough to still be JITtable, in which case we would match with the JITted version of $B with interpreted version of $A, instead of failing, right?

THanks.
Previous: Junio C HamanoNext: Junio C Hamano
Message 16 of 36 in “grep: fall back to interpreter mode if JIT fails”
  1. grep: fall back to interpreter mode if JIT failsMathias Krause, Dec 16, 2022
  2. Ævar Arnfjörð BjarmasonDec 16, 2022
  3. Mathias KrauseDec 16, 2022
  4. Junio C HamanoDec 16, 2022
  5. Carlo ArenasDec 17, 2022
  6. Ævar Arnfjörð BjarmasonDec 19, 2022
  7. Mathias KrauseDec 20, 2022
  8. Ævar Arnfjörð BjarmasonDec 20, 2022
  9. Mathias KrauseJan 18, 2023
  10. Ævar Arnfjörð BjarmasonJan 18, 2023
  11. Mathias KrauseJan 19, 2023
  12. Junio C HamanoDec 16, 2022
  13. Mathias KrauseDec 20, 2022
  14. grep: fall back to interpreter if JIT memory allocation failsMathias Krause, Jan 27, 2023
  15. Junio C HamanoJan 27, 2023
  16. Junio C HamanoJan 27, 2023
  17. Junio C HamanoJan 27, 2023
  18. Mathias KrauseJan 29, 2023
  19. Mathias KrauseJan 29, 2023
  20. Junio C HamanoJan 29, 2023
  21. Ævar Arnfjörð BjarmasonJan 30, 2023
  22. Junio C HamanoJan 30, 2023
  23. Ævar Arnfjörð BjarmasonJan 31, 2023
  24. Mathias KrauseJan 30, 2023
  25. Junio C HamanoJan 30, 2023
  26. Junio C HamanoJan 30, 2023
  27. Junio C HamanoJan 30, 2023
  28. Ramsay JonesJan 30, 2023
  29. Junio C HamanoJan 30, 2023
  30. Mathias KrauseJan 31, 2023
  31. Junio C HamanoJan 31, 2023
  32. Mathias KrauseJan 31, 2023
  33. Mathias KrauseJan 31, 2023
  34. Mathias KrauseJan 29, 2023
  35. grep: fall back to interpreter if JIT memory allocation failsMathias Krause, Jan 31, 2023
  36. Junio C HamanoJan 31, 2023

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.