Hi,
I am working on getting libosmocom package into Fedora, there is review request in [1].
During the packaging I found following minor issues: - Incorrect FSF addresses in source code, e.g. there is written in the sources: 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
But the correct address is (as taken from the https://www.gnu.org/licenses/gpl-2.0.html): 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA
- exit call in the library: /usr/lib64/libosmovty.so.3.0.0 exit@GLIBC_2.2.5 libraries shouldn't generally call exit.
thanks & regards
Jaroslav
Hi Jaroslav,
On Tue, Nov 10, 2015 at 04:56:23AM -0500, Jaroslav Skarvada wrote:
I am working on getting libosmocom package into Fedora, there is review request in [1].
Thanks for this!
During the packaging I found following minor issues:
- Incorrect FSF addresses in source code
This is easy to fix. Do you already have a patch to fix all occurrences? I'd happily apply that.
- exit call in the library: /usr/lib64/libosmovty.so.3.0.0 exit@GLIBC_2.2.5
libraries shouldn't generally call exit.
libosmovty started as a fork of the VTY (telnet command line interface) code of GNU zebra. So the code was not written as a library to begin with, but was part of zebra itself.
In any case, addressing those issues is not particularly easy, as said functions should never fail at this point, but they don't have a way to return an error code, or their callers simply assume they always succeed.
So the best we can do without breaking API and ABI (and modifying all users of the library) is to change the exit(1) calls into an assert. Would that be better?
----- Original Message -----
Hi Jaroslav,
On Tue, Nov 10, 2015 at 04:56:23AM -0500, Jaroslav Skarvada wrote:
I am working on getting libosmocom package into Fedora, there is review request in [1].
Thanks for this!
During the packaging I found following minor issues:
- Incorrect FSF addresses in source code
This is easy to fix. Do you already have a patch to fix all occurrences? I'd happily apply that.
Patch attempting to fix it is attached.
- exit call in the library: /usr/lib64/libosmovty.so.3.0.0 exit@GLIBC_2.2.5
libraries shouldn't generally call exit.
libosmovty started as a fork of the VTY (telnet command line interface) code of GNU zebra. So the code was not written as a library to begin with, but was part of zebra itself.
In any case, addressing those issues is not particularly easy, as said functions should never fail at this point, but they don't have a way to return an error code, or their callers simply assume they always succeed.
So the best we can do without breaking API and ABI (and modifying all users of the library) is to change the exit(1) calls into an assert. Would that be better?
Nice, thanks. Please note, this is not something that blocks Fedora review, I just wanted to point it out, e.g. to consider it for next API update sometimes in the future
thanks & regards
Jaroslav
Hi Jaroslav,
On Wed, Nov 11, 2015 at 10:23:11AM -0500, Jaroslav Skarvada wrote:
During the packaging I found following minor issues:
- Incorrect FSF addresses in source code
Patch attempting to fix it is attached.
Patch applied, thanks.
- exit call in the library: /usr/lib64/libosmovty.so.3.0.0 exit@GLIBC_2.2.5
libraries shouldn't generally call exit.
So the best we can do without breaking API and ABI (and modifying all users of the library) is to change the exit(1) calls into an assert. Would that be better?
Nice, thanks. Please note, this is not something that blocks Fedora review, I just wanted to point it out, e.g. to consider it for next API update sometimes in the future
Thanks. I've just applied a patch to change the exit() calls into OSMO_ASSERT() instead, which is a better and more generic way to do this anyway.
The VTY code doesn't get much love from us, we simply adapted it from Zebra back in the early days of OpenBSC, and it sits there without much review/improvement/change.
Regards, Harald
12.11.2015 13:50, Harald Welte пишет:
The VTY code doesn't get much love from us, we simply adapted it from Zebra back in the early days of OpenBSC, and it sits there without much review/improvement/change.
Speaking of which - upstream (quagga, successor to zebra) would welcome if anyone could abstract their vty code and split it into separate library.
See https://bugzilla.quagga.net/show_bug.cgi?id=847 for details.
cheers, Max.
-
Harald Welte -
Jaroslav Skarvada -
Suraev