This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH v2 5/5] Post-cleanup 2: minimize _G_config.h.
On 06/02/2018 18:17, Zack Weinberg wrote:
> On Tue, Feb 6, 2018 at 2:06 PM, Adhemerval Zanella
> <adhemerval.zanella@linaro.org> wrote:
>> On 05/02/2018 21:35, Zack Weinberg wrote:
>>> Nearly everything in _G_config.h is either junk or more appropriately
>>> defined elsewhere:
> ...
>>
>> LGTM in general with just _IO_HAVE_ST_BLKSIZE suggestion below.
> ...
>>> -#if _IO_HAVE_ST_BLKSIZE
>>> +#if defined _STATBUF_ST_BLKSIZE
>>> if (st.st_blksize > 0 && st.st_blksize < _IO_BUFSIZ)
>>> size = st.st_blksize;
>>> #endif
>>
>> I think it is more logical to have _IO_HAVE_ST_BLKSIZE defined as 0
>> on generic bits/stat.h and 1 for Linux bits/stat.h (which is also
>> the way to we are aiming to handle these kind of macros).
>
> _STATBUF_ST_BLKSIZE, with #undef/1 semantics, has been part of our
> sys/stat.h for a long time; it doesn't have a _lot_ of external users,
> but it has a few (see
> https://codesearch.debian.net/search?q=_STATBUF_ST_BLKSIZE). I am
> reluctant to complicate this patchset by messing with it. Is this a
> hard objection?
I do not think this is a patch blocker.
>
>>> +/* Setting this macro to 1 enables the use of the _vtable_offset bias
>>> + in _IO_JUMPS_FUNCS, below. This is only needed for new-format
>>> + _IO_FILE in libc that must support old binaries (see oldfileops.c). */
>>
>> Ok, this was indeed a convoluted check (I had to write it down to realize
>> _G_IO_NO_BACKWARD_COMPAT here is superflous).
>
> I want to thank Joseph for insisting on the "no changes to installed
> stripped libraries" standard for this patchset; I got this change
> wrong initially and I'm not sure I would have noticed that it was
> wrong if I wasn't being suspicious of any change to the generated code.
>
> zw
>