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

Re: [PATCH 01/10] environment: move access to core.maxTreeDepth into repo settings

From
Patrick Steinhardt <ps@pks.im>
Date
Jan 12, 2026, 09:21 UTC
Message-ID
<aWS9JLb3UyxOwg9Z@pks.im>
In-Reply-To
<20260109213021.2546-2-l.s.r@web.de>
On Fri, Jan 09, 2026 at 10:30:12PM +0100, René Scharfe wrote:
Show 35 quoted lines
> diff --git a/git-compat-util.h b/git-compat-util.h
> index b0673d1a450..bebcf9f698c 100644
> --- a/git-compat-util.h
> +++ b/git-compat-util.h
> @@ -578,6 +578,30 @@ static inline bool strip_suffix(const char *str, const char *suffix,
>  #define DEFAULT_PACKED_GIT_LIMIT \
>  	((1024L * 1024L) * (size_t)(sizeof(void*) >= 8 ? (32 * 1024L * 1024L) : 256))
>  
> +#ifdef _MSC_VER
> +  /*
> +   * When traversing into too-deep trees, Visual C-compiled Git seems to
> +   * run into some internal stack overflow detection in the
> +   * `RtlpAllocateHeap()` function that is called from within
> +   * `git_inflate_init()`'s call tree. The following value seems to be
> +   * low enough to avoid that by letting Git exit with an error before
> +   * the stack overflow can occur.
> +   */
> +#define DEFAULT_MAX_ALLOWED_TREE_DEPTH 512
> +#elif defined(GIT_WINDOWS_NATIVE) && defined(__clang__) && defined(__aarch64__)
> +  /*
> +   * Similar to Visual C, it seems that on Windows/ARM64 the clang-based
> +   * builds have a smaller stack space available. When running out of
> +   * that stack space, a `STATUS_STACK_OVERFLOW` is produced. When the
> +   * Git command was run from an MSYS2 Bash, this unfortunately results
> +   * in an exit code 127. Let's prevent that by lowering the maximal
> +   * tree depth; This value seems to be low enough.
> +   */
> +#define DEFAULT_MAX_ALLOWED_TREE_DEPTH 1280
> +#else
> +#define DEFAULT_MAX_ALLOWED_TREE_DEPTH 2048
> +#endif
> +
>  int git_open_cloexec(const char *name, int flags);
>  #define git_open(name) git_open_cloexec(name, O_RDONLY)
>  

Why is this new constant defined in "git-compat-util.h"? That feels way too global for my taste for a value that should ultimately only be required by "repo-settings.c". Is there a reason why it's not declared in "repo-settings.h"?

I think we could even get away with declaring it in "repo-settings.c". The expectation is that the settings need to be prepared anyway, so we can simply set up the default when parsing the config.

Patrick
Previous: René ScharfeNext: René Scharfe
Message 18 of 25 in “tree: stop using the_repository”
  1. 00/10 tree: stop using the_repositoryRené Scharfe, Jan 9, 2026
  2. 03/10 add-interactive: use repo_parse_tree_indirect()René Scharfe, Jan 9, 2026
  3. 04/10 bloom: use repo_parse_tree()René Scharfe, Jan 9, 2026
  4. 08/10 tree: use repo_parse_tree()René Scharfe, Jan 9, 2026
  5. Patrick SteinhardtJan 12, 2026
  6. 09/10 tree: stop using the_repositoryRené Scharfe, Jan 9, 2026
  7. Patrick SteinhardtJan 12, 2026
  8. Junio C HamanoJan 12, 2026
  9. Patrick SteinhardtJan 12, 2026
  10. Junio C HamanoJan 12, 2026
  11. Junio C HamanoJan 12, 2026
  12. Patrick SteinhardtJan 12, 2026
  13. René ScharfeJan 12, 2026
  14. Patrick SteinhardtJan 13, 2026
  15. 10/10 cocci: convert parse_tree functions to repo_ variantsRené Scharfe, Jan 9, 2026
  16. 02/10 tree: add repo_parse_tree*()René Scharfe, Jan 9, 2026
  17. 01/10 environment: move access to core.maxTreeDepth into repo settingsRené Scharfe, Jan 9, 2026
  18. Patrick SteinhardtJan 12, 2026
  19. René ScharfeJan 12, 2026
  20. 07/10 path-walk: use repo_parse_tree_gently()René Scharfe, Jan 9, 2026
  21. 05/10 delta-islands: use repo_parse_tree()René Scharfe, Jan 9, 2026
  22. 06/10 pack-bitmap-write: use repo_parse_tree()René Scharfe, Jan 9, 2026
  23. 11/10 cocci: remove obsolete the_repository rulesRené Scharfe, Jan 15, 2026
  24. Patrick SteinhardtJan 16, 2026
  25. Junio C HamanoJan 16, 2026

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.