On Wed, Jun 4, 2014 at 5:21 PM, H. Peter Anvin <hpa at zytor.com> wrote:> On 06/04/2014 11:09 AM, Raphael S Carvalho wrote: >> Before proceeding with the MultiFS patchset, let's discuss the MultiFS syntax. >> >> It's currently as follows: >> (hd[disk number]:[partition number])/path/to/file >> >> Any objections? >> > > As others have pointed out, I'd rather use a comma, but even more > specifically, we already *have* a very sophisticated syntax for > partition selection: this is what chain.c32 currently consumes. It lets > us specify partitions not just by disk/partition number but also by > various kinds of other attributes like MBR signatures and UUIDs. > > So we should just (re-)use that syntax.Agreed. the whole chain directory will have to be moved to com32/lib, or common code pulled into a new file. Which one does sound better to you?> > -hpa > >-- Raphael S. Carvalho
On Wed, Jun 4, 2014 at 5:36 PM, Raphael S Carvalho <raphael.scarv at gmail.com> wrote:> On Wed, Jun 4, 2014 at 5:21 PM, H. Peter Anvin <hpa at zytor.com> wrote: >> On 06/04/2014 11:09 AM, Raphael S Carvalho wrote: >>> Before proceeding with the MultiFS patchset, let's discuss the MultiFS syntax. >>> >>> It's currently as follows: >>> (hd[disk number]:[partition number])/path/to/file >>> >>> Any objections? >>> >> >> As others have pointed out, I'd rather use a comma, but even more >> specifically, we already *have* a very sophisticated syntax for >> partition selection: this is what chain.c32 currently consumes. It lets >> us specify partitions not just by disk/partition number but also by >> various kinds of other attributes like MBR signatures and UUIDs. >> >> So we should just (re-)use that syntax. > Agreed. the whole chain directory will have to be moved to com32/lib, > or common code pulled into a new file. Which one does sound better > to you?Although the syntax parser (in option.c) doesn't assume the parentheses neither the path post the closed parentheses, it's also bloated and seems not to fit well the MultiFS purposes. On the other hand, find_dp and its helper functions at chain.c could be nicely re-used with a few changes.>> >> -hpa >> >> > > > > -- > Raphael S. Carvalho-- Raphael S. Carvalho
It would be good to clean it up so out can be used for both. Just teach it that ) ends the parse. On June 4, 2014 2:06:22 PM PDT, Raphael S Carvalho <raphael.scarv at gmail.com> wrote:>On Wed, Jun 4, 2014 at 5:36 PM, Raphael S Carvalho ><raphael.scarv at gmail.com> wrote: >> On Wed, Jun 4, 2014 at 5:21 PM, H. Peter Anvin <hpa at zytor.com> wrote: >>> On 06/04/2014 11:09 AM, Raphael S Carvalho wrote: >>>> Before proceeding with the MultiFS patchset, let's discuss the >MultiFS syntax. >>>> >>>> It's currently as follows: >>>> (hd[disk number]:[partition number])/path/to/file >>>> >>>> Any objections? >>>> >>> >>> As others have pointed out, I'd rather use a comma, but even more >>> specifically, we already *have* a very sophisticated syntax for >>> partition selection: this is what chain.c32 currently consumes. It >lets >>> us specify partitions not just by disk/partition number but also by >>> various kinds of other attributes like MBR signatures and UUIDs. >>> >>> So we should just (re-)use that syntax. >> Agreed. the whole chain directory will have to be moved to com32/lib, >> or common code pulled into a new file. Which one does sound better >> to you? >Although the syntax parser (in option.c) doesn't assume the parentheses >neither the path post the closed parentheses, it's also bloated and >seems >not to fit well the MultiFS purposes. > >On the other hand, find_dp and its helper functions at chain.c could be >nicely re-used with a few changes. > >>> >>> -hpa >>> >>> >> >> >> >> -- >> Raphael S. Carvalho-- Sent from my mobile phone. Please pardon brevity and lack of formatting.
On Wed, Jun 4, 2014 at 6:06 PM, Raphael S Carvalho <raphael.scarv at gmail.com> wrote:> On Wed, Jun 4, 2014 at 5:36 PM, Raphael S Carvalho > <raphael.scarv at gmail.com> wrote: >> On Wed, Jun 4, 2014 at 5:21 PM, H. Peter Anvin <hpa at zytor.com> wrote: >>> On 06/04/2014 11:09 AM, Raphael S Carvalho wrote: >>>> Before proceeding with the MultiFS patchset, let's discuss the MultiFS syntax. >>>> >>>> It's currently as follows: >>>> (hd[disk number]:[partition number])/path/to/file >>>> >>>> Any objections? >>>> >>> >>> As others have pointed out, I'd rather use a comma, but even more >>> specifically, we already *have* a very sophisticated syntax for >>> partition selection: this is what chain.c32 currently consumes. It lets >>> us specify partitions not just by disk/partition number but also by >>> various kinds of other attributes like MBR signatures and UUIDs. >>> >>> So we should just (re-)use that syntax. >> Agreed. the whole chain directory will have to be moved to com32/lib, >> or common code pulled into a new file. Which one does sound better >> to you? > Although the syntax parser (in option.c) doesn't assume the parentheses > neither the path post the closed parentheses, it's also bloated and seems > not to fit well the MultiFS purposes.We could though adapt opt_parse_args in chain/options.c to work with MultiFS syntax (hda,b)/path/, it currently works only for hda,b. Something like: s = malloc(len = strlen()); memcpy(s, path, len); if (s[0] == '(') s++; if (closed_p = strtol(s, ')') *closed_p = '\0';> > On the other hand, find_dp and its helper functions at chain.c could be > nicely re-used with a few changes. > >>> >>> -hpa >>> >>> >> >> >> >> -- >> Raphael S. Carvalho > > > > -- > Raphael S. Carvalho-- Raphael S. Carvalho