cygport i686/x86 library build undefined type in sys/stat.h

cygport i686/x86 library build undefined type in sys/stat.h

[Brian Inglis]

Trying to build latest libssh2 1.9.0 using cygport, works under x86_64,
but under i686/x86 fails with:

In file included from */usr/i686-pc-cygwin/include/sys/stat.h*:22,
                  from ...:
/usr/include/cygwin/stat.h:27:3: error: unknown type name ‘timestruc_t’
    27 |   timestruc_t   st_atim;
       |   ^~~~~~~~~~~

The equivalent include path to */usr/i686-pc-cygwin/include/sys/stat.h*
does not exist under 64 bit as /usr/x86_64-pc-cygwin/include,
and cygcheck -f does not come up with a hit.

The only obvious work around the timeframe of these files appears to be a
local test newlib-cygwin build and install for format_proc_cpuinfo and
format_proc_swaps testing:

$ head -v /proc/version
==> /proc/version <==
CYGWIN_NT-10.0-19042-WOW64 version 3.2.1-340.i686 (...@...) (gcc version 10.2.0 (GCC) ) 2021-04-30 12:55 UTC

Has anyone any idea, from seeing anything similar happen on their installs,
where this directory and these files may have come from,
what may have installed them, and whether it is likely to be safe to
rename or remove them?

-- 
Take care. Thanks, Brian Inglis, Calgary, Alberta, Canada

This email may be disturbing to some readers as it contains
too much technical detail. Reader discretion is advised.
[Data in binary units and prefixes, physical quantities in SI.]

Re: cygport i686/x86 library build undefined type in sys/stat.h

[Marco Atzeri]

On 15.05.2021 06:16, Brian Inglis wrote:
> Trying to build latest libssh2 1.9.0 using cygport, works under x86_64,
> but under i686/x86 fails with:
> 
> In file included from */usr/i686-pc-cygwin/include/sys/stat.h*:22,
>                   from ...:
> /usr/include/cygwin/stat.h:27:3: error: unknown type name ‘timestruc_t’
>     27 |   timestruc_t   st_atim;
>        |   ^~~~~~~~~~~
> 
> The equivalent include path to */usr/i686-pc-cygwin/include/sys/stat.h*
> does not exist under 64 bit as /usr/x86_64-pc-cygwin/include,
> and cygcheck -f does not come up with a hit.
> 

It is defined under:

/usr/include/machine/types.h

#ifndef __timestruc_t_defined
#define __timestruc_t_defined
typedef struct timespec timestruc_t;
#endif /*__timestruc_t_defined*/


check that the <threads.h> is included

$ cat prova.c
#include <threads.h>
int main(){
timestruc_t abstime;
}

$ gcc -c  prova.c
$ ls -l prova.o
-rw-r--r-- 1 Marco Kein 656 May 15 07:59 prova.o

Re: cygport i686/x86 library build undefined type in sys/stat.h

[Marco Atzeri]

On 15.05.2021 08:01, Marco Atzeri wrote:
> On 15.05.2021 06:16, Brian Inglis wrote:
>> Trying to build latest libssh2 1.9.0 using cygport, works under x86_64,
>> but under i686/x86 fails with:
>>
>> In file included from */usr/i686-pc-cygwin/include/sys/stat.h*:22,
>>                   from ...:
>> /usr/include/cygwin/stat.h:27:3: error: unknown type name ‘timestruc_t’
>>     27 |   timestruc_t   st_atim;
>>        |   ^~~~~~~~~~~
>>
>> The equivalent include path to */usr/i686-pc-cygwin/include/sys/stat.h*
>> does not exist under 64 bit as /usr/x86_64-pc-cygwin/include,
>> and cygcheck -f does not come up with a hit.
>>
> 
> It is defined under:
> 
> /usr/include/machine/types.h
> 
> #ifndef __timestruc_t_defined
> #define __timestruc_t_defined
> typedef struct timespec timestruc_t;
> #endif /*__timestruc_t_defined*/
> 
> 
> check that the <threads.h> is included

of course <time.h> is a more natural choice

Re: cygport i686/x86 library build undefined type in sys/stat.h

[Ken Brown]

On 5/15/2021 12:16 AM, Brian Inglis wrote:
> Trying to build latest libssh2 1.9.0 using cygport, works under x86_64,
> but under i686/x86 fails with:
> 
> In file included from */usr/i686-pc-cygwin/include/sys/stat.h*:22,
>                   from ...:
> /usr/include/cygwin/stat.h:27:3: error: unknown type name ‘timestruc_t’
>     27 |   timestruc_t   st_atim;
>        |   ^~~~~~~~~~~
> 
> The equivalent include path to */usr/i686-pc-cygwin/include/sys/stat.h*
> does not exist under 64 bit as /usr/x86_64-pc-cygwin/include,
> and cygcheck -f does not come up with a hit.
> 
> The only obvious work around the timeframe of these files appears to be a
> local test newlib-cygwin build and install for format_proc_cpuinfo and
> format_proc_swaps testing:

It looks like something went wrong with the install of your test build.  In 
particular, the headers went to the wrong place.

> $ head -v /proc/version
> ==> /proc/version <==
> CYGWIN_NT-10.0-19042-WOW64 version 3.2.1-340.i686 (...@...) (gcc version 10.2.0 
> (GCC) ) 2021-04-30 12:55 UTC
> 
> Has anyone any idea, from seeing anything similar happen on their installs,
> where this directory and these files may have come from,
> what may have installed them, and whether it is likely to be safe to
> rename or remove them?

Not only is it safe, but I would say it's mandatory.  You should also look for 
other remnants of your test install and remove them.  In general, I think it's a 
bad idea to do test installs in the same Cygwin installation that you use for 
building packages.

Ken