[PATCH v3 2/2] x86_64/i386 fenv: Replace symlink with include fenv_stub.c

Joel Sherrill joel@rtems.org
Thu Feb 20 13:54:00 GMT 2020


On Thu, Feb 20, 2020 at 3:01 AM Corinna Vinschen <vinschen@redhat.com>
wrote:

> Hey Joel,
>
> the patch itself is fine, but I'm puzzeling over the commit message.
>

That was the end result of the investigation into the user's issue. I had
to be
pinged privately to even notice that my fenv addition of symlinks triggered
the
failure. It just wasn't obvious at first to me. It was hard to see how the
fenv patch
could trigger this.

My conclusion is that avoiding symlinks in source distributions falls into
the
same rule bucket as not using capitalization to distinguish files in the
same
directory. Works almost everywhere not not quite. Just like we used to avoid
long file names.


>
> On Feb 19 08:03, Joel Sherrill wrote:
> > Having symlinks for these files led to an issue reported to the
> > RTEMS Project that showed up on MSYS2. Per the report we received,
>
> MSYS2 is a Cygwin clone which understands any kind of Cygwin symlink,
> an invention from way back when Windows really didn't have symlinks.
>
> > there are no such things as links on Windows,
>
> NTFS has symlinks since Windows Vista, about 2007.
>

Yeah. It is an issue with tar. Apparently you can't create symlinks
to non-existent files at least with bsdtar. If the link is in the tar file
before
the real file, it fails to create the file. Maybe GNU tar covers this
up.

>
> > it is a copy of the
> > file and if it does not exist it has to fail. The working all depends
> > on the order the tar file is built. If the links appear in the tar file
> > before the source exists, it cannot copy the file.
>
> Isn't that just somebody using an inadequate tool here?  Do you have a
> pointer to a bugzilla/github/whatever issue clarifying the actual
> problem?
>

Here is the thread. It is a few messages in before investigation shows it
is a tar file issue.

https://lists.rtems.org/pipermail/devel/2020-February/057184.html

The RTEMS ticket is here:

https://devel.rtems.org/ticket/3868

I know this is an odd one.

--joel

>
>
> Thanks,
> Corinna
>
> --
> Corinna Vinschen
> Cygwin Maintainer
> Red Hat
>



More information about the Newlib mailing list