[Pkg-owncloud-maintainers] FTBS ocsync & qtkeychain

[Svante Signell]

On Thu, 2013-09-19 at 12:42 +0200, Pino Toscano wrote:
> Hi,
> 
> In data giovedì 19 settembre 2013 12:16:49, Svante Signell ha scritto:
> > The patch is attached. Additionally, the freebsd patch is modified to
> > include errno.h for definition of ENODATA and EPIPE. The compiler
> > complains that EPIPE is not defined otherwise. Maybe this file can be
> > included unconditionally, without harm. I wonder how the error codes
> > are defined for freebsd if errno.h is not included?
> 
> Reviews below.
> 
> Storing statically the result of getenv is wrong, since it points to the 
> global environment array which can be reallocated anytime.

A solution would be to remove the static keyword then.

> Also, no need to rename envp to home, makes the change bigger for no 
> reason.

Agreed!

> Furthermore, your changes are losing the check for an empty HOME.

I don't think so, in that case getenv returns NULL 

> A better change is the following, which is easier:

I thought of doing something like that, but did not. You beat me to it, thanks!

> No need to misuse __GNU__ for this, errno.h is a standard and such it 
> can be included anytime. In case it poses issues on Windows (sigh), you 
> can include it in the else of the _WIN32 block (as people in -bsd@ 
> suggested regarding fcntl.h).

See my comment above! The errno.h inclusion place was under discussion.

> >  #ifdef _WIN32
> >  #include <windef.h>
> >  #include <winbase.h>
> > @@ -50,6 +54,11 @@
> >  #define geteuid() 0
> >  #endif
> >  
> > +#ifndef ENODATA
> > +#define ENODATA EPIPE
> > +#endif
> 
> This may be problematic. c_copy in src/std/c_file.c uses ENODATA as 
> errno in a couple of error situations, but since c_copy is an exported 
> symbol (although its header is not installed), users of this function 
> would need to do this ENODATA→EPIPE mapping themselves too.

This is not part of my patch. Address this to upstream/debian maintainer
directly.