[OE-core] [PATCH V2] psmisc: Remove including sys/user.h and __WORDSIZE

Khem Raj raj.khem at gmail.com
Wed Mar 30 23:29:32 UTC 2016


> On Mar 30, 2016, at 3:53 PM, Phil Blundell <pb at pbcl.net> wrote:
> 
> On Wed, 2016-03-30 at 17:56 +0000, Khem Raj wrote:
>> +Subject: [PATCH] Use UINTPTR_MAX instead of __WORDSIZE
>> +
>> +Do not include sys/user.h since it conflicts with
>> +pt_regs struct from kernel APIs in asm/ptrace.h
>> +
>> +Signed-off-by: Khem Raj <raj.khem at gmail.com>
> 
> This description (and the subject line of the commit) implies that you
> are removing the include unconditionally.  But actually, you are only
> removing it for non-glibc platforms.

OK.

> 
> Also, the change itself seems slightly backwards since the ptrace
> manpage suggests that <sys/user.h> is actually the right header to be
> using for these definitions and possibly it's <asm/ptrace.h> that should
> be eliminated.  Maybe that's not feasible here but it would be nice if
> the patch header explained why.  And…

The linux ptrace(2) manpage is written with glibc as the only reference. Unfortunately, its not a POSIX API
so there is no standard per se. The motivation for this patch to look for a middle ground, between
different libc implementaitons.

musl’s sys/user.h defines elements that conflict with asm/ptrace.h especially on ppc, may be musl community
should take a closer look at what can be done since many tracing utilities e.g. strace gdb etc. rely on
this, may be its worth a libc item there.

> 
>> -SRC_URI += "file://0001-Typo-in-fuser-makes-M-on-all-the-time.patch \
>> -            file://0002-Include-limits.h-for-PATH_MAX.patch \
>> -            "
>> +SRC_URI = "${SOURCEFORGE_MIRROR}/psmisc/psmisc-${PV}.tar.gz \
>> +           file://0001-Typo-in-fuser-makes-M-on-all-the-time.patch \
>> +           file://0002-Include-limits.h-for-PATH_MAX.patch \
>> +           file://0001-Use-UINTPTR_MAX-instead-of-__WORDSIZE.patch \
>> +           "
> 
> ... this change is not mentioned in the commit message.  Why have you
> added the tarball to SRC_URI directly (rather than letting the .inc do
> that as it did before)?
> 
> p.
> 
> 

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 204 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://lists.openembedded.org/pipermail/openembedded-core/attachments/20160330/e0d02b9e/attachment-0002.sig>


More information about the Openembedded-core mailing list