Re: [PATCH] doc: mention environ pitfall

[Eric Blake]

On 05/31/2018 08:54 AM, Eric Blake wrote:
> Based on a bug report to the Cygwin list:
> https://cygwin.com/ml/cygwin/2018-05/msg00321.html
>
> * doc/posix-functions/environ.texi (environ): Assigning NULL to
> environ is a glibc extension.
>
> Signed-off-by: Eric Blake <address@hidden>
> ---
>
> Does anyone have a better list of platforms where environ=NULL
> actually fails, or should I just leave this text as written here?

No answer, so I pushed the patch.

>   ChangeLog                        | 6 ++++++
>   doc/posix-functions/environ.texi | 6 ++++++
>   2 files changed, 12 insertions(+)
>
> diff --git a/ChangeLog b/ChangeLog
> index b23b10919..5aaa8b708 100644
> --- a/ChangeLog
> +++ b/ChangeLog
> @@ -1,3 +1,9 @@
> +2018-05-31  Eric Blake  <address@hidden>
> +
> +       doc: mention environ pitfall
> +       * doc/posix-functions/environ.texi (environ): Assigning NULL to
> +       environ is a glibc extension.
> +
>   2018-05-27  Colin Watson  <address@hidden>
>
>         bootstrap: document source fetching in --help
> diff --git a/doc/posix-functions/environ.texi b/doc/posix-functions/environ.texi
> index a6c00958d..6d1b8cfaa 100644
> --- a/doc/posix-functions/environ.texi
> +++ b/doc/posix-functions/environ.texi
> @@ -34,4 +34,10 @@ environ
>   @item
>   The address of this variable is not a compile-time constant on some platforms:
>   mingw.
> address@hidden
> +Assigning NULL to @code{environ} to clear all variables is not
> +portable; better is to assign @code{environ} to one-element array
> +containing a NULL pointer.  That said, an empty environment is not
> +portable either, as some systems require particular environment
> +variables to be present to operate consistently.
>   @end itemize

--
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3266
Virtualization:  qemu.org | libvirt.org