Re: [PATCH v3 00/17] Catch syscall group

[Pedro Alves <palves at redhat dot com>]

On 04/26/2015 02:25 AM, Gabriel Krisman Bertazi wrote:
> Hello,
> 
> This is the third version of the catch syscall group patches.  This
> version presents the following modifications, other than fixing the
> things Sergio and Doug pointed out in previous reviews:
> 
> * Created syscall groups for every architecture supported by catch
>   syscall.
> 
> The previous version only implemented syscalls for amd64.  I used a
> script to generate the xmls, and based the group field information on
> strace, so please share your thoughts if you disagree with any group.

Thanks.

OOC, does strace also maintain a separate group table per architecture?
Was your script something like a map of syscall _name_ to group?

I was wondering if we couldn't share most of the grouping
per-architecture, e.g., by having each arch syscall file xi:include a
base Linux default groups file, that listed the grouping without
the syscall number.  E.g., create a linux-defaults.xml like:

 <syscall-defaults name="select" groups="descriptor"/>
 ...
 <syscall-defaults name="openat" groups="descriptor,file"/>

And then the arch-specific syscall files would do:

 <xi:include href="linux-defaults.xml"/>
 ...
 <syscall name="select" number="82"/>
 ...
 <syscall name="arch_specific_foo" number="200" groups="descriptor"/>
 ...

And then we teach gdb about syscall-default-groups, or
we could even instead preprocess the architecture xml to expand
the groups into each syscall with xsltproc (we already use
this tool, see gdb/features/Makefile and gdb/features/number-regs.xsl).

> 
> I babysitted the xml generation scripts but I was not able to actually
> test them on architectures other than amd64 and i386.  It would be very
> nice if someone could actually run catch-syscall.exp on other
> architectures.

Thanks,
Pedro Alves