Pending patches & other changes

Pending patches & other changes

[Rich Felker]

Here is a list of patches I'm aware of that have been sent to the list
that I haven't yet applied or rejected:

[PATCH] generate debug_frame info for __syscall_cp_asm (i386) so gdb can get backtrace
[PATCH] fmtmsg: verify that label is in the correct format.
[PATCH] replace 'hlt' by 'ud2' in i386/x32/x86_64 a_crash implementations
[PATCH] conforming strverscmp() implementation

There are also a few patches from nsz on the radar:

http://port70.net/~nsz/musl/0001-add-IPTOS_CLASS_-macros-to-netinet-ip.h.patch
http://port70.net/~nsz/musl/0001-complex-fix-ctanh-0-i-nan-and-ctanh-0-i-inf.patch
http://port70.net/~nsz/musl/0001-inet_pton-clean-up.patch

and the BRE extensions that are still under discussion, and possibly
some cleanup of some always-false null checks in the failure path of
regcomp.

There's also the idea of changing getlogin[_r] to probe the actual
controlling terminal and its owner to get the "logged in user" rather
than just using $LOGNAME. I've done some work on this but it's not
done and not clean for commit.

Anything else I'm missing? Most of this probably won't go in the
release but I at least want to make sure we have a good inventory of
what's still pending.

Rich

Re: Pending patches & other changes

[Szabolcs Nagy]

* Rich Felker <dalias@libc.org> [2015-04-27 23:48:20 -0400]:
> Here is a list of patches I'm aware of that have been sent to the list
> that I haven't yet applied or rejected:
> 
> [PATCH] generate debug_frame info for __syscall_cp_asm (i386) so gdb can get backtrace
> [PATCH] fmtmsg: verify that label is in the correct format.
> [PATCH] replace 'hlt' by 'ud2' in i386/x32/x86_64 a_crash implementations
> [PATCH] conforming strverscmp() implementation
> 
> There are also a few patches from nsz on the radar:
> 
> http://port70.net/~nsz/musl/0001-add-IPTOS_CLASS_-macros-to-netinet-ip.h.patch
> http://port70.net/~nsz/musl/0001-complex-fix-ctanh-0-i-nan-and-ctanh-0-i-inf.patch
> http://port70.net/~nsz/musl/0001-inet_pton-clean-up.patch
> 
> and the BRE extensions that are still under discussion, and possibly
> some cleanup of some always-false null checks in the failure path of
> regcomp.
> 
> There's also the idea of changing getlogin[_r] to probe the actual
> controlling terminal and its owner to get the "logged in user" rather
> than just using $LOGNAME. I've done some work on this but it's not
> done and not clean for commit.
> 
> Anything else I'm missing? Most of this probably won't go in the
> release but I at least want to make sure we have a good inventory of
> what's still pending.
> 

i forgot to do this one in elf.h:

-#define SHF_EXCLUDE         (1 << 31)
+#define SHF_EXCLUDE         (1U << 31)

(there is no other invalid <<31)

Re: Pending patches & other changes

[Rich Felker]

On Tue, Apr 28, 2015 at 11:43:21AM +0200, Szabolcs Nagy wrote:
> * Rich Felker <dalias@libc.org> [2015-04-27 23:48:20 -0400]:
> > Here is a list of patches I'm aware of that have been sent to the list
> > that I haven't yet applied or rejected:
> > 
> > [PATCH] generate debug_frame info for __syscall_cp_asm (i386) so gdb can get backtrace
> > [PATCH] fmtmsg: verify that label is in the correct format.
> > [PATCH] replace 'hlt' by 'ud2' in i386/x32/x86_64 a_crash implementations
> > [PATCH] conforming strverscmp() implementation
> > 
> > There are also a few patches from nsz on the radar:
> > 
> > http://port70.net/~nsz/musl/0001-add-IPTOS_CLASS_-macros-to-netinet-ip.h.patch
> > http://port70.net/~nsz/musl/0001-complex-fix-ctanh-0-i-nan-and-ctanh-0-i-inf.patch
> > http://port70.net/~nsz/musl/0001-inet_pton-clean-up.patch
> > 
> > and the BRE extensions that are still under discussion, and possibly
> > some cleanup of some always-false null checks in the failure path of
> > regcomp.
> > 
> > There's also the idea of changing getlogin[_r] to probe the actual
> > controlling terminal and its owner to get the "logged in user" rather
> > than just using $LOGNAME. I've done some work on this but it's not
> > done and not clean for commit.
> > 
> > Anything else I'm missing? Most of this probably won't go in the
> > release but I at least want to make sure we have a good inventory of
> > what's still pending.
> > 
> 
> i forgot to do this one in elf.h:
> 
> -#define SHF_EXCLUDE         (1 << 31)
> +#define SHF_EXCLUDE         (1U << 31)
> 
> (there is no other invalid <<31)

This was discussed quite a bit on libc-alpha (glibc list) and it seems
the consensus there is that changing the type to unsigned is
potentially dangerous and it would be better to use (-0x7fffffff-1) or
perhaps ((1<<30) * -2) if it's more clear. What do you think?

Rich

Re: Pending patches & other changes

[Szabolcs Nagy]

* Rich Felker <dalias@libc.org> [2015-04-28 09:46:47 -0400]:
> On Tue, Apr 28, 2015 at 11:43:21AM +0200, Szabolcs Nagy wrote:
> > i forgot to do this one in elf.h:
> > 
> > -#define SHF_EXCLUDE         (1 << 31)
> > +#define SHF_EXCLUDE         (1U << 31)
> > 
> > (there is no other invalid <<31)
> 
> This was discussed quite a bit on libc-alpha (glibc list) and it seems
> the consensus there is that changing the type to unsigned is
> potentially dangerous and it would be better to use (-0x7fffffff-1) or
> perhaps ((1<<30) * -2) if it's more clear. What do you think?
> 

the glibc bikeshed ended with florian saying

"I thought some more about this, and have changed my opinion completely.
 Making the constant unsigned is less risky than making it negative
 because of potential sign extension issues."
https://sourceware.org/ml/libc-alpha/2015-04/msg00264.html

and they committed the original 1U<<31 patch today

Re: Pending patches & other changes

[Rich Felker]

On Mon, Apr 27, 2015 at 11:48:20PM -0400, Rich Felker wrote:
> Here is a list of patches I'm aware of that have been sent to the list
> that I haven't yet applied or rejected:
> 
> [PATCH] generate debug_frame info for __syscall_cp_asm (i386) so gdb can get backtrace
> [PATCH] fmtmsg: verify that label is in the correct format.
> [PATCH] replace 'hlt' by 'ud2' in i386/x32/x86_64 a_crash implementations
> [PATCH] conforming strverscmp() implementation
> 
> There are also a few patches from nsz on the radar:
> 
> http://port70.net/~nsz/musl/0001-add-IPTOS_CLASS_-macros-to-netinet-ip.h.patch
> http://port70.net/~nsz/musl/0001-complex-fix-ctanh-0-i-nan-and-ctanh-0-i-inf.patch

I applied these two.

> http://port70.net/~nsz/musl/0001-inet_pton-clean-up.patch

I suspect this should be done but I want to review it well and haven't
gotten around to that yet. Let's look again after release.

> and the BRE extensions that are still under discussion, and possibly

Since adding these now would be a big committment to keep them I think
we should hold off until/unless there's evidence that they're widely
needed. I'd prefer getting apps fixed but if it's a recurring headache
for users maybe we should just add them.

> some cleanup of some always-false null checks in the failure path of
> regcomp.

This is just cleanup and I want to be sure it's correct before doing
it, so let's look again after release.

Rich