This is the mail archive of the gdb-patches@sourceware.org mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [patch#2 3/6] set auto-load local-gdbinit warn-and-*


On Fri, 30 Mar 2012 09:25:41 +0200, Eli Zaretskii wrote:
> >  If enabled, canned sequences of commands are loaded when debugger starts\n\
> >  from .gdbinit file in current directory.  Such files are deprecated,\n\
> >  use script associated with inferior executable file instead.\n\
>    ^^^^^^^^^^
> Either "use a script" or "use scripts", probably the former.

Use "use a script" (it is in [patch 2/X]).


> > +This feature is deprecated, please use @ref{objfile-gdb.rc file} instead;
> 
> "objfile" should be in @var, and it should explain what is "objfile".

Added:

This feature is deprecated, please use @ref{@var{objfile}-gdb.gdb file}
instead.  The canned sequence of commands will be then specific for loaded
@var{objfile}---typically the main executable---after converting an init file in
the current directory into the @var{objfile}-gdb.gdb file..


> >  gdb-scripts:  Canned sequences of commands auto-loading is on.
> >  libthread-db:  Inferior specific libthread_db auto-loading is on.
> > -local-gdbinit:  Current directory .gdbinit script auto-loading is on.
> > +local-gdbinit:  Current directory .gdbinit script auto-loading is yes.
> 
> Why "yes"?

parse_binary_operation accepts any of 'on', '1', 'yes' or 'enable' (and sure
off/0/no/disable).  It then displays the boolean only as "on" and "off".

Somehow it seems more appropriate for these auto-loading settings to use
yes/no instead of on/off.  This is also why I use yes/no in the documentation.
When I added the warning option I had to choose only one pair, therefore
I chose yes/no.  But this means the default displayed boolean values are new
yes/no (and not on/off).  It also means it no longer accepts the alternative
words on/1/enable (and off/0/disable); but that is not a real regression when
the setting never existed before this patchset.

There are two possibilities:
 * Introduce new option some "set local-gdbinit-warning <boolean>" as
   suggested by Doug IIUC and technically keep everything boolean.
 * Switch the words used in documentation from yes/no to on/off, therefore
   also make the switch on/off/warn-and-on/warn-and-off.



Thanks,
Jan


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]