You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Vladimir Prus <gh...@cs.msu.su> on 2002/02/14 14:16:19 UTC

[PATCH] Updating a locally modified file with endline translation

Log message:
Fixes a bug that prevented updating a locally modified file when endline 
translation for the file is turned on.

    * subversion/libsvn_wc/get_editor.c (svn_wc_install_file): Make path
          for a temporary file relative to the directory where update happens.

    * subversion/tests/clients/cmdline/update_tests.py 
          (update_modified_with_translation): New function.

Patch (against rev 1272) 

Index: ./subversion/libsvn_wc/get_editor.c
===================================================================
--- ./subversion/libsvn_wc/get_editor.c
+++ ./subversion/libsvn_wc/get_editor.c	Thu Feb 14 17:04:57 2002
@@ -1843,6 +1843,13 @@
                                                         FALSE,
                                                         pool));
 
+                      /* Make the temporary working file name relative to 
+                         the parent directory. */
+                      tmp_working = svn_stringbuf_ncreate
+                        (tmp_working->data + parent_dir->len + 1,
+                         tmp_working->len - parent_dir->len - 1,
+                         pool);
+
                       /* Copy the working file to tmp-working with
                          LF's, and any keywords contracted. */
 
Index: ./subversion/tests/clients/cmdline/update_tests.py
===================================================================
--- ./subversion/tests/clients/cmdline/update_tests.py
+++ ./subversion/tests/clients/cmdline/update_tests.py	Thu Feb 14 17:04:12 
2002
@@ -474,6 +474,48 @@
                                                None, None, None, None, 0,
                                                '-r', '0')
 
+#----------------------------------------------------------------------
+
+def update_modified_with_translation(sbox):
+  "ensure update for locally modified file with eol translation works"
+
+  if sbox.build():
+    return 1
+
+  wc_dir = sbox.wc_dir
+  cwd = os.getcwd()
+  os.chdir(wc_dir)
+
+  # Create initial content of rho and turn on newline translation.
+  rho_path = os.path.join('A', 'D', 'G', 'rho')
+  f = open(rho_path, "w")
+  f.write("1\n2\n3\n4\n5\n6\n7\n8\n9\n")
+  f.close()
+  svntest.main.run_svn(None, 'propset', 'svn:eol-style', 'native', rho_path)
+  svntest.main.run_svn(None, 'ci', '-m', '"log msg"', rho_path)  
+
+  # Make immeditate revision
+  f = open(rho_path, "w")
+  f.write("1\n2\n3\n4\n4.5\n5\n6\n7\n8\n9\n")
+  f.close()
+  svntest.main.run_svn(None, 'ci', '-m', '"log msg"', rho_path)
+
+  # Locally modify a file
+  f = open(rho_path, "w")
+  f.write("1\n2\n3\n4\n4.5\n5\n6\n7\n8\n9\n10\n")
+  f.close()
+
+  # Try updating, and see if anything is printed to stderr
+  result = 0
+  if svntest.main.run_svn(None, 'up', '-r1')[1]:
+    result = 1
+
+  os.chdir(cwd)
+
+  return result
+  
+
+
 
 ########################################################################
 # Run the tests
@@ -485,6 +527,7 @@
               update_binary_file_2,
               update_ignores_added,
               update_to_rev_zero,
+              update_modified_with_translation,
               # update_missing,
              ]
 
A question:
In the test code above, I have to explicitly change directory to working 
copy. Otherwise, all commits/updates fill happen in subversion tree. However, 
other tests in update_tests.py don't do that? What am I missing?

- Volodya

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: [PATCH] Updating a locally modified file with endline translation

Posted by Ben Collins-Sussman <su...@collab.net>.
cmpilato@collab.net writes:

> Vladimir Prus <gh...@cs.msu.su> writes:
> 
> > A question:
> > In the test code above, I have to explicitly change directory to
> > working copy. Otherwise, all commits/updates fill happen in
> > subversion tree. However, other tests in update_tests.py don't do
> > that? What am I missing?
> 
> Most other tests, when building paths to operate on, use:
> 
>    os.path.join (wc_dir, 'A', 'D', 'G', 'rho')
> 
> instead of what you did:
> 
>    os.path.join ('A', 'D', 'G', 'rho')
> 

Also, a side comment: you're running 'commit' and other subcommands
directly via the run_svn() routine.  This is legitimate, but it
doesn't provide as much coverage as most of the other tests.  The
other tests use a different API: run_and_verify_FOO(), where foo is
one of {commit, update, status}.

Admittedly, this is a harder API to use -- but it's a lot more
thorough.  It scans the contents of the working copy, it parses the
output of the subcommand, and it parses of the output of 'svn st -v'
after the subcommand is finished.  For people who have been mystified
about writing tests this way, read the huge doc comment at the top of
subversion/tests/clients/cmdline/svntest/tree.py.



---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: [PATCH] Updating a locally modified file with endline translation

Posted by cm...@collab.net.
Vladimir Prus <gh...@cs.msu.su> writes:

> A question:
> In the test code above, I have to explicitly change directory to
> working copy. Otherwise, all commits/updates fill happen in
> subversion tree. However, other tests in update_tests.py don't do
> that? What am I missing?

Most other tests, when building paths to operate on, use:

   os.path.join (wc_dir, 'A', 'D', 'G', 'rho')

instead of what you did:

   os.path.join ('A', 'D', 'G', 'rho')

The way our testing setup current exists, it is impossible to test any
Subversion operation that needs `.' as the target without doing the
cwd stuff like you did.  If your test is not expected to behave any
differently in:

   (cd foo; svn <command> .)

versus:

   svn <command> foo

You should probably just prepend wc_dir to all the paths in use.



---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: [PATCH] Updating a locally modified file with endline translation

Posted by Karl Fogel <kf...@newton.ch.collab.net>.
Vladimir Prus <gh...@cs.msu.su> writes:
> I don't see any problem here: the second assignment uses the result of the 
> first, doesn't it? This is similar to what happens on lines 1784
> and 1808 in the same file.

Indeed -- that makes sense, I misread your code.  Thanks!

> > Also it seems odd that the old way would work in most circumstances,
> > yet not for your reproduction recipe.  
> 
> Does the old way work indeed? Is there any existing test that passes the 
> block of code that I've patched? 

Perhaps not, since its translation specific, and we've deferred those
tests till the alpha phase.

I'll reexamine with your points in mind,
-K

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: [PATCH] Updating a locally modified file with endline translation

Posted by Vladimir Prus <gh...@cs.msu.su>.
Karl Fogel wrote:

> > ===================================================================
> > --- ./subversion/libsvn_wc/get_editor.c
> > +++ ./subversion/libsvn_wc/get_editor.c	Thu Feb 14 17:04:57 2002
> > @@ -1843,6 +1843,13 @@
> >                                                          FALSE,
> >                                                          pool));
> >
> > +                      /* Make the temporary working file name relative
> > to +                         the parent directory. */
> > +                      tmp_working = svn_stringbuf_ncreate
> > +                        (tmp_working->data + parent_dir->len + 1,
> > +                         tmp_working->len - parent_dir->len - 1,
> > +                         pool);
> > +
> >                        /* Copy the working file to tmp-working with
> >                           LF's, and any keywords contracted. */
>
> The context here doesn't show it, but this new assignment to
> tmp_working immediately follows another assignment (by reference) to
> tmp_working.  They can't both be right, since the first *use* of the
> variable occurs later on. :-)

I don't see any problem here: the second assignment uses the result of the 
first, doesn't it? This is similar to what happens on lines 1784  and 1808 in 
the same file.

> Also it seems odd that the old way would work in most circumstances,
> yet not for your reproduction recipe.  

Does the old way work indeed? Is there any existing test that passes the 
block of code that I've patched? 

- Volodya

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: [PATCH] Updating a locally modified file with endline translation

Posted by Karl Fogel <kf...@newton.ch.collab.net>.
Vladimir Prus <gh...@cs.msu.su> writes:
> Log message:
> Fixes a bug that prevented updating a locally modified file when endline 
> translation for the file is turned on.
> 
>     * subversion/libsvn_wc/get_editor.c (svn_wc_install_file): Make path
>           for a temporary file relative to the directory where update happens.
> 
>     * subversion/tests/clients/cmdline/update_tests.py 
>           (update_modified_with_translation): New function.
> 
> Patch (against rev 1272) 
> 
> Index: ./subversion/libsvn_wc/get_editor.c
> ===================================================================
> --- ./subversion/libsvn_wc/get_editor.c
> +++ ./subversion/libsvn_wc/get_editor.c	Thu Feb 14 17:04:57 2002
> @@ -1843,6 +1843,13 @@
>                                                          FALSE,
>                                                          pool));
>  
> +                      /* Make the temporary working file name relative to 
> +                         the parent directory. */
> +                      tmp_working = svn_stringbuf_ncreate
> +                        (tmp_working->data + parent_dir->len + 1,
> +                         tmp_working->len - parent_dir->len - 1,
> +                         pool);
> +
>                        /* Copy the working file to tmp-working with
>                           LF's, and any keywords contracted. */

The context here doesn't show it, but this new assignment to
tmp_working immediately follows another assignment (by reference) to
tmp_working.  They can't both be right, since the first *use* of the
variable occurs later on. :-)

Also it seems odd that the old way would work in most circumstances,
yet not for your reproduction recipe.  Suspect there's something
deeper going on here, and am looking...

-K

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org