Alexander Skwar wrote on Wed, Aug 04, 2010 at 21:23:50 +0200:
> Hi.
>
> 2010/8/4 Daniel Shahaf <d...@daniel.shahaf.name>
>
> > No. That file is auto-generated from libsvn_fs_fs/*.sql (IIRC, by
> > build/transform_sql.py).
> >
> > (And, apparently, the autogeneration mechanism has changed, since in trunk
> > I get
> >
> > % grep REP_CACHE_DB_SQL .../subversion/libsvn_fs_fs/*h
> > src/svn/trunk.d/subversion/libsvn_fs_fs/rep-cache-db.h:#define
> > REP_CACHE_DB_SQL_DECLARE_STATEMENTS(varname) \
> >
> > .)
>
> Hm. I ran transform_sql.py manually like so:
>
> cd build
> ./transform_sql.py ../subversion/libsvn_fs_fs/rep-cache-db.sql foo.h2
>
> or
> ~/.software/Python-2.5.2/bin/python2.5 transform_sql.py
> ../subversion/libsvn_fs_fs/rep-cache-db.sql foo.h3
>
> or
> /opt/csw/bin/python2.5 transform_sql.py
> ../subversion/libsvn_fs_fs/rep-cache-db.sql foo.h4
>
> or
> /usr/bin/python2.4 transform_sql.py
> ../subversion/libsvn_fs_fs/rep-cache-db.sql foo.h4
>
> In all these cases, the generated "foo.*" file is fine.
>
> Ah!
>
> --($ ~/Source/subversion-1.6.12/build)-- grep -r transform_sql ..
> ../build/transform_sql.py:# transform_sql.py -- create a header file
> with the appropriate SQL variables
> ../build/generator/gen_win.py: import transform_sql
> ../build/generator/gen_win.py: transform_sql.main(open(sql + '.sql',
> 'r'),
> ../Makefile.in: $(top_srcdir)/build/transform_sql.sh $< < $< >
> $(top_srcdir)/$@
> ../Makefile: $(top_srcdir)/build/transform_sql.sh $< < $< >
> $(top_srcdir)/$@
>
> It's *NOT* the Python .py file which is run, but the .sh Shell
> script!
>
> And it's this script, which doesn't work!
>
> --($ ~/Source/subversion-1.6.12/build)-- /bin/sh -x ./transform_sql.sh
> ../subversion/libsvn_fs_fs/rep-cache-db.sql
> + basename ../subversion/libsvn_fs_fs/rep-cache-db.sql
> + + trtr [a-z]. _+
> [A-Z]
> tr - _
> varname=rep_cZche_db_sql
> […]
>
> What you can see here, is that varname is already wrong!
>
> From the sh file:
>
> varname=`basename $1 | tr "[a-z]" "[A-Z]" | tr "-" "_" | tr "." "_"`
>
> Okay. It's using "tr", which is:
> --($ ~/Source/subversion-1.6.12/build)-- which tr
> /usr/xpg6/bin/tr
>
> And now pay attention:
>
> --($ ~/Source/subversion-1.6.12/build)--
> E=../subversion/libsvn_fs_fs/rep-cache-db.sql
>
> --($ ~/Source/subversion-1.6.12/build)-- basename $E
> rep-cache-db.sql
>
> --($ ~/Source/subversion-1.6.12/build)-- basename $E | tr "[a-z]" "[A-Z]"
> rep-cZche-db.sql
>
> And now I use GNU tr:
>
> --($ ~/Source/subversion-1.6.12/build)-- basename $E | gtr "[a-z]" "[A-Z]"
> REP-CACHE-DB.SQL
>
> Hmmmm.....
>
> It's not a GNUism. If I use /usr/bin/tr instead of /usr/xpg6/bin/tr or
> /usr/xpg4/bin/tr, then it also works.
>
Could we use the tr that configure finds?
> Seeing that "tr" doesn't always work the same way, maybe it would
> be useful to use something else? Eg. perl, which seems to be the most
> "portable"?
>
> --($ ~/Source/subversion-1.6.12/build)-- date | sed 'y/[a-z]/[A-Z]/'
> Mittwoch, 4. August 2010 21:10 Uhr CEST
>
> -> sed didn't work
>
How about 'sed -e s/a/A/ -e s/b/B/ -e ... -e s/z/Z/'?
> --($ ~/Source/subversion-1.6.12/build)-- date | /usr/bin/awk '{print
> toupper($0)}'
> Mittwoch, 4. August 2010 21:09 Uhr CEST
>
> -> awk also didn't work
>
> But perl works.
>
Right now, we don't require Perl (at all) and we require Python only from
developers (but not from folks who download our tarballs).
> --($ ~/Source/subversion-1.6.12/build)-- date | perl -pe 'tr/a-z/A-Z/; '
> MITTWOCH, 4. AUGUST 2010 21:21 UHR CEST
>
> And it would allow to combine the seperate tr commands, eg.:
>
> varname=`basename $1 | perl -pe 'tr/a-z/A-Z/; tr/-/_/; tr/./_/;'`
>
> Does it make sense to file this as a bug?
>
Yes, feel free to file a bug and/or summarize the issue on d...@.
Daniel
(see also
http://subversion.apache.org/docs/community-guide/general.html#patches)
> cheers,
>
> Alexander
> --
> ↯ Lifestream (Twitter, Blog, …) ↣ http://alexs77.soup.io/ ↯
> ↯ Chat (Jabber/Google Talk) ↣ a.sk...@gmail.com , AIM: alexws77 ↯
