Greg Stein wrote:
> On Sat, Jan 24, 2009 at 14:53, Hyrum K. Wright
> <hyrum_wright_at_mail.utexas.edu> wrote:
>> Greg Stein wrote:
>>> On Fri, Jan 23, 2009 at 23:20, Hyrum K. Wright <hyrum_at_hyrumwright.org>
>>>> For efficiency reasons, the system has converted the large body of this
>>>> message into an attachment.
>>>> ---------- Forwarded message ----------
>>>> From: hwright_at_tigris.org
>>>> To: svn_at_subversion.tigris.org
>>>> Date: Fri, 23 Jan 2009 14:20:25 -0800
>>>> Subject: svn commit: r35440 - branches/explore-wc/subversion/libsvn_wc
>>>> Author: hwright
>>>> Date: Fri Jan 23 14:20:24 2009
>>>> New Revision: 35440
>>>> Exile non-wc-ng entries reading code to another file, so we can pare down
>>>> entries.c at our leisure.
>>>> This duplicates a bunch of code, but lets the new file stand on its own.
>>>> After completely migrating entries.c, we can determine which bits are
>>>> duplicated and refactor then.
>>>> * subversion/libsvn_wc/old-and-busted.c:
>>>> * subversion/libsvn_wc/entries.c
>>>> (entries_accumulator, handle_start_tag, parse_entries_xml,
>>>> write_entry_xml, write_entries_xml): Remove.
>>>> (svn_wc__entries_write): Don't worry about writing xml entries.
>>>> branches/explore-wc/subversion/libsvn_wc/old-and-busted.c (contents,
>>>> props changed)
>>> Why is there still reading code in entries.c? I don't understand why
>>> you don't just call read_entries_old() ?
>> My plan is to start paring it away as it gets replaced. Eventually, there
>> won't be any of the reading code in entries.c.
> You mean that you'll read from both entries *and* sqlite, filling in
> wc_entry preferably from sqlite? And remove the code to fill in entry
> from entries?
> So upgrading the format becomes: read_entries_old() followed by
> write_entries() (which writes to both) ?
> Am I understanding correctly? If not, then can you explain in a bit
> more detail? I'm not getting it?
All this move did was make a copy of the old reading code, so that I can hack
away at entries.c at will. I'm not too worried about the details of upgrading
the wc right now, but having the existing reading code tucked away in it's own
little file will make upgrading easier when the time arrives.
My plan is something like:
1 Write stuff into the sqlite database, alongside writing to the entries file
2 Read said stuff from the sqlite database, alongside reading from the entries
3 Remove redundant code reading and writing to the entries file
4 Lather/Rinse/Repeat for the various bits of the entries file until
everything is migrated over
This is just my way of doing things incrementally and making sure I don't break
things hopelessly and spend inordinate amounts of time trying to track down the
breakage. The move of the existing read code to old-and-busted.c is just an
effort to facilitate the plan, and to make sure I don't go slashing stuff which
we'll need down the road.
Does that make it a bit clearer?
Received on 2009-01-24 19:08:48 CET