Re: obstack changes: relicense under LGPLv2+

Bruno Haible via Gnulib discussion list writes: > 2025-05-05 Bruno Haible > > obstack: Relicense under LGPLv2+. > Alan Modra's permission is at > . > Paul Eggert's permission is at >

[PATCH 10/19] * lib/obstack.in.h: Move include.

This is merely to make the glibc reviewer’s job easier, since the include location doesn’t matter. --- lib/obstack.in.h | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/lib/obstack.in.h b/lib/obstack.in.h index 0f2677dba4..f53b78c15c 100644 --- a/lib/obstack.in.h +++ b/lib/

[PATCH 08/19] obstack: rename to _OBSTACK_CHUNK_SIZE_T

* lib/obstack.in.h (_OBSTACK_CHUNK_SIZE_T): Rename from _CHUNK_SIZE_T, for namespace prefix consistency. --- ChangeLog| 4 lib/obstack.in.h | 6 +++--- 2 files changed, 7 insertions(+), 3 deletions(-) diff --git a/ChangeLog b/ChangeLog index ed1c57575b..771309c182 100644 --- a/Change

[PATCH 02/19] obstack: stop depending on alignof

* lib/obstack.c (__alignof__) [!__GNUC__]: Default to alignof, not to alignof_type. * modules/obstack (Depends-on): Depend on alignasof directly, not indirectly via alignof. --- ChangeLog | 6 ++ lib/obstack.c | 3 +-- modules/obstack | 2 +- 3 files changed, 8 insertions(+), 3 deletio

[PATCH 09/19] obstack: size_t → _OBSTACK_CHUNK_SIZE_T

* lib/obstack.c (_obstack_begin, _obstack_begin_1): * lib/obstack.in.h (struct obstack, obstack_init, obstack_begin) (obstack_specify_allocation) (obstack_specify_allocation_with_arg, obstack_chunkfun): Use _OBSTACK_CHUNK_SIZE_T instead of size_t in places where glibc uses unsigned long rather than

[PATCH 05/19] obstack: define PTR_INT_TYPE

* lib/obstack.in.h (PTR_INT_TYPE): Define, for compatibility with glibc even though we don’t need the type for anything other than compatibility. Include stddef.h only if needed, like glibc obstack.h does. --- ChangeLog| 6 ++ lib/obstack.in.h | 15 +-- 2 files changed, 1

[PATCH 06/19] obstack: use presence of uintptr_t

The code depends on the presence of uintptr_t, not on whether ptrdiff_t is narrower than a pointer, so use C macros that reflect this. Also, clean up the namespace in the typical GCC-like case * lib/obstack.in.h (_OBSTACK_UINTPTR_TYPE): New macro. Include only if needed (i.e., not GCC-like). (__PT

[PATCH 11/19] obstack: use glibc’s temp slot names

* lib/obstack.in.h (struct obstack): Rename temp fields back to what glibc uses, to avoid any glibc concerns about source code compatibility. All uses changed back. This affect only non-GNU-compatible compilers. --- ChangeLog| 6 ++ lib/obstack.in.h | 52

[PATCH 12/19] obstack: _obstack_free → __obstack_free

* lib/obstack.in.h (__obstack_free): Rename back from _obstack_free. This is for compatibility with glibc, which in turn is for compatibility with old Gnulib. All uses changed. --- ChangeLog| 5 + lib/obstack.c| 4 +++- lib/obstack.in.h | 10 ++ 3 files changed, 14 inser

[PATCH 15/19] libc-config, obstack: Oracle __extension__ support

* lib/cdefs.h, lib/obstack.in.h (__extension__): Oracle Developer Studio 12.6 also supports __extension__. --- ChangeLog| 4 lib/cdefs.h | 2 +- lib/obstack.in.h | 3 ++- 3 files changed, 7 insertions(+), 2 deletions(-) diff --git a/ChangeLog b/ChangeLog index dd0825acdf..c0691c

[PATCH 17/19] obstack: use signed chunk sizes

* lib/obstack.in.h (_OBSTACK_CHUNK_SIZE_T) [!__GL_GNULIB_HEADER]: Now long, not unsigned long, since that’s what glibc does. --- ChangeLog| 4 lib/obstack.in.h | 2 +- 2 files changed, 5 insertions(+), 1 deletion(-) diff --git a/ChangeLog b/ChangeLog index 8967f3d54f..eb0ba5d166 1006

[PATCH 18/19] obstack: simplify default size, alignment

* lib/obstack.c (MAX, DEFAULT_ALIGNMENT, DEFAULT_ROUNDING): Remove. (_obstack_begin_worker): Use __alignof__ (max_align_t) instead of calculating it by hand, possibly incorrectly on oddball platforms. Estimate a good size of 4000 bytes rather than using a no-longer-valid value. * modules/obstack (

[PATCH 13/19] obstack: be more like glibc in return values

This partly reverts commit d15b2da0ac25e085ce30a9e2672624999ce910a6 dated 2014-11-03. It keeps the part of that commit that fixed the return type of non-gcc versions of obstack_make_room, obstack_grow, obstack_grow0, obstack_1grow, obstack_ptr_grow_fast, obstack_int_grow_fast, obstack_blank so tha

[PATCH 14/19] obstack: be more like glibc for __extension__

* lib/obstack.in.h (__extension__): Use glibc-like syntax for testing gcc version. --- ChangeLog| 4 lib/obstack.in.h | 3 +-- 2 files changed, 5 insertions(+), 2 deletions(-) diff --git a/ChangeLog b/ChangeLog index 1e83db3d5e..dd0825acdf 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1

[PATCH 19/19] obstack: match obstack.c with glibc

Make this file easier to compare with glibc obstack.c. * lib/obstack.c: Include shlib-compat.h or libc-config.h. Move some !_OBSTACK_NO_ERROR_HANDLER declarations up. Add _LIBC-only compatibility decls. * modules/obstack (Depends-on): Add libc-config. --- ChangeLog | 7 +++ lib/obstack.

[PATCH 16/19] obstack: fix undefined behavior if big alignment

* lib/obstack.c: Include stddef.h, for offsetof. (align_size_up): New static function. (_obstack_begin_worker, _obstack_newchunk): Ensure that the chunk limit is aligned, too. (_obstack_begin_worker): Don’t assume that 4096 is large enough, as the alignment could be larger. * lib/obstack.in.h (obst

[PATCH 01/19] obstack: fix unlikely buffer overrun in glibc

This bug is not plausible on any Gnulib platform, but it could happen in glibc if the requested size is UINT_MAX. * lib/obstack.c (_obstack_newchunk): Fail if requested size is 0. * lib/obstack.in.h (obstack_grow0): Check for room <= len, not room <= len + 1 where len + 1 could overflow. --- Chang

[PATCH 04/19] obstack: change license wording

Change license wording to match glibc, to ease future merging. --- lib/obstack.c| 19 ++- lib/obstack.in.h | 19 ++- 2 files changed, 20 insertions(+), 18 deletions(-) diff --git a/lib/obstack.c b/lib/obstack.c index 7d19f6bd3a..92e2a1206f 100644 --- a/lib/obst

[PATCH 03/19] obstack: merge better with glibc

When compiled for glibc, keep the fake array content size at 4, for API compatibility with older glibc. * lib/obstack.in.h (__FLEXIBLE_ARRAY_MEMBER): Remove. (_OBSTACK_CHUNK_CONTENTS_SIZE): New macro. (struct _obstack_chunk): Use it for contents size. * modules/obstack (Depends-on): Add flexmember.

[PATCH 07/19] obstack: use int where glibc does

Make it easier to merge with glibc, by using int where the glibc API uses int, when compiling for glibc. This does not affect Gnulib behavior. * lib/obstack.in.h (_OBSTACK_INDEX_T): New macro. * lib/obstack.in.h (struct obstack): * lib/obstack.c (_obstack_begin_worker, obstack_begin) (_obstack_begi

[PATCH 00/19] prepare to merge Gnulib obstack back into glibc

Prompted by the recent attempt[1] to fix an obstack-related bug[2] in glibc, a bug that was already fixed in Gnulib, I have attempted to make the Gnulib code usable in glibc without changing the glibc ABI and with the most-innocuous changes to the undocumented part of the glibc API, and installed t

Re: New module windows-cygpath

The native Windows 'git' program prints file names in the form C:/Users/foo/bar . (It is linked with MSYS2, not with Cygwin.) We don't need to call cygpath in this case, but we still need to convert C:/Users/foo/bar -> C:\Users\foo\bar like 'cygpath -w' would do. 2025-05-05 Bruno Haible

Re: vc-mtime: Add API for more efficient use of git

> vc-mtime: Add API for more efficient use of git. Two more patches are needed, to make it work on native Windows. 2025-05-05 Bruno Haible vc-mtime: Another fix for native Windows. * lib/vc-mtime.c (SLASH): New macro. (max_vc_mtime): Use it as expected directory

New module windows-cygpath

The module 'cygpath' is used to convert file names in a Cygwin program before passing them to a native Windows program. The opposite situation is to convert file names in a native Windows program that were received from a Cygwin program. This needs a different module. 2025-05-05 Bruno Haible

Re: Improving support for find_shared_library_fullname (module 'relocatable')

On Fri, 2 May 2025 at 02:30, Bruno Haible wrote: I don't think requiring -ldl is a "small" price to pay. It increases the > startup times of the programs linked to that library. Therefore how about > using dladdr only on those platforms where it is in libc? This would cover > all the platform tha

Re: obstack changes: relicense under LGPLv2+

Alan Modra wrote: > > Would you please relicense your changes under LGPLv2+? That is, give us > > the ability to relicense the 'obstack' module under LGPLv2+ and then be > > able to flow back the changes into glibc? > > OK by me. Thank you! I'm implementing the license change: 2025-05-05 Bruno

Re: obstack changes: relicense under LGPLv2+

On Sun, May 04, 2025 at 12:10:45PM +0200, Bruno Haible wrote: > Would you please relicense your changes under LGPLv2+? That is, give us > the ability to relicense the 'obstack' module under LGPLv2+ and then be > able to flow back the changes into glibc? OK by me. -- Alan Modra

Re: obstack changes: relicense under LGPLv2+

On 2025-05-04 03:10, Bruno Haible wrote: Would you please relicense your changes under LGPLv2+? That is, give us the ability to relicense the 'obstack' module under LGPLv2+ and then be able to flow back the changes into glibc? Yes, of course. And I don't think things were unclear at all, as th

Re: Testsuite failure in (unprivileged) user namespace

Hi Collin, Collin Funk writes: > Ludovic Courtès writes: > >>> I am only a relatively new user of guix, so I can't speak for which >>> approach would be preferable from their perspective. I CC'ed Ludovic >>> since he implemented the unprivileged guix daemon approach. >> >> I’m not sure how ‘mod