git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH v2 6/6] t9109-git-svn-props.sh: split up several pipes

From
Matthew DeVore <matvore@google.com>
Date
Sep 19, 2018, 02:56 UTC
Message-ID
<CAMfpvhJ3ye_7LWaQ1abXKtMB=O1sfOz6xYN=7acrVdLOksq9eA@mail.gmail.com>
In-Reply-To
<CAPig+cT5BLu2onbuTBbZ_mMzNMkEuPk5-g2d5YKw4V6Z42Y3aQ@mail.gmail.com>
On Mon, Sep 17, 2018 at 6:57 PM Eric Sunshine <sunshine@sunshineco.com> wrote:
Show 10 quoted lines
>
> On Mon, Sep 17, 2018 at 6:25 PM Matthew DeVore <matvore@google.com> wrote:
> > t9109-git-svn-props.sh: split up several pipes
>
> Similar to my comment about 5/6, this title talks about the mechanical
> changes made by the patch but not the intent. Perhaps reword it like
> this:
>
>     t9109: avoid swallowing Git exit code upstream of a pipe
>
Here is my new commit description:
    t9109: don't swallow Git errors upstream of pipe
    'git ... | foo' will mask any errors or crashes in git, so split up such
    pipes.
    One testcase uses several separate pipe sequences in a row which are
    awkward to split up. Wrap the split-up pipe in a function so the
    awkwardness is not repeated.
    Signed-off-by: Matthew DeVore <matvore@google.com>
Show 10 quoted lines
> > diff --git a/t/t9101-git-svn-props.sh b/t/t9101-git-svn-props.sh
> > @@ -190,16 +190,21 @@ EOF
> > +# Note we avoid using pipes in order to ensure that git exits with 0.
>
> This new comment doesn't really add value for someone reading the
> patch without knowing the history leading up to the point the comment
> was added. It should probably be dropped. (The actual text of the
> comment is rather confusing anyhow since avoiding pipes has nothing to
> do with ensuring that git exits with 0, thus another reason why this
> comment ought to be dropped.)

Yes, this comment was worded quite poorly. I've removed it since I agree it doesn't add a lot of value.

Show 9 quoted lines
>
> >  test_expect_success 'test propget' "
> > -       git svn propget svn:ignore . | cmp - prop.expect &&
> > +       test_propget () {
> > +               git svn propget $1 $2 >observed
>
> The &&-chain is broken here, which means you're losing the exit status
> from the Git command anyhow (despite the point of the patch being to
> avoid losing it).
Fixed.
>
> Also, for consistency, how about calling this "actual" rather than "observed"?
Done.
Show 10 quoted lines
>
> > +               cmp - $3
>
> This is just wrong. The "-" argument to 'cmp' says to read from
> standard input, but there is nothing being passed to 'cmp' on standard
> input anymore now that you're removed the pipe. I'm guessing that you
> really meant to use "observed" here (and reverse the order of
> arguments to be consistent with the expect-then-actual idiom).
> Finally, since these (apparently) might be binary, you can use
> test_cmp_bin() instead.

Fixed, except for the test_cmp_bin part. My understanding is that git svn propget is supposed to be printing human-readable strings. My understanding is based soley on this page: http://svnbook.red-bean.com/en/1.6/svn.ref.svn.c.propget.html

Show 26 quoted lines
>
> > +       } &&
> > +       test_propget svn:ignore . prop.expect &&
> >         cd deeply &&
> > -       git svn propget svn:ignore . | cmp - ../prop.expect &&
> > -       git svn propget svn:entry:committed-rev nested/directory/.keep \
> > -         | cmp - ../prop2.expect &&
> > -       git svn propget svn:ignore .. | cmp - ../prop.expect &&
> > -       git svn propget svn:ignore nested/ | cmp - ../prop.expect &&
> > -       git svn propget svn:ignore ./nested | cmp - ../prop.expect &&
> > -       git svn propget svn:ignore .././deeply/nested | cmp - ../prop.expect
> > +       test_propget svn:ignore . ../prop.expect &&
> > +       test_propget svn:entry:committed-rev nested/directory/.keep \
> > +               ../prop2.expect &&
> > +       test_propget svn:ignore .. ../prop.expect &&
> > +       test_propget svn:ignore nested/ ../prop.expect &&
> > +       test_propget svn:ignore ./nested ../prop.expect &&
> > +       test_propget svn:ignore .././deeply/nested ../prop.expect
> >         "
>
> After this patch, the test is even more broken than appears at first
> glance since the test body is inside double-quotes. This means that
> the $1, $2, $3 inside the test_propget() function are getting expanded
> _before_ the function itself is ever defined, to whatever bogus values
> $1, $2, $3 hold at that point. I can't see how this could ever have
> worked (except only appearing to work by pure accident).
Fixed, and here is the new test:
test_expect_success 'test propget' "
        test_propget () {
                git svn propget \$1 \$2 >actual &&
                cmp \$3 actual
        } &&
        test_propget svn:ignore . prop.expect &&
        cd deeply &&
        test_propget svn:ignore . ../prop.expect &&
        test_propget svn:entry:committed-rev nested/directory/.keep \
                ../prop2.expect &&
        test_propget svn:ignore .. ../prop.expect &&
        test_propget svn:ignore nested/ ../prop.expect &&
        test_propget svn:ignore ./nested ../prop.expect &&
        test_propget svn:ignore .././deeply/nested ../prop.expect
        "

I confirmed that git's exit code is being checked by putting "!" in front of git and making sure the test failed. I made sure that "actual" was actually being compared and the exit code of "cmp" was checked by adding "echo foo >> actual &&" before cmp, and again making sure the test failed. This test should be well-formed now.

Previous: Eric SunshineNext: Eric Sunshine
Message 64 of 66 in “Cleanup tests for test_cmp argument ordering and "|" placement”
  1. 0/2 Cleanup tests for test_cmp argument ordering and "|" placementMatthew DeVore, Sep 15, 2018
  2. 1/2 t/*: fix pipe placement and remove \'sMatthew DeVore, Sep 15, 2018
  3. Jonathan NiederSep 17, 2018
  4. Matthew DeVoreSep 17, 2018
  5. 2/2 t/*: fix ordering of expected/observed argumentsMatthew DeVore, Sep 15, 2018
  6. Matthew DeVoreSep 17, 2018
  7. Junio C HamanoSep 15, 2018
  8. 0/6 Clean up tests for test_cmp arg ordering and pipe placementMatthew DeVore, Sep 17, 2018
  9. 4/6 tests: Add linter check for pipe placement styleMatthew DeVore, Sep 17, 2018
  10. Eric SunshineSep 18, 2018
  11. Matthew DeVoreSep 19, 2018
  12. 0/5 Clean up tests for test_cmp arg ordering and pipe placementMatthew DeVore, Sep 21, 2018
  13. 1/5 CodingGuidelines: add shell piping guidelinesMatthew DeVore, Sep 21, 2018
  14. Eric SunshineSep 21, 2018
  15. Matthew DeVoreSep 21, 2018
  16. SZEDER GáborSep 24, 2018
  17. Matthew DeVoreSep 25, 2018
  18. SZEDER GáborSep 27, 2018
  19. Matthew DeVoreOct 1, 2018
  20. 2/5 tests: standardize pipe placementMatthew DeVore, Sep 21, 2018
  21. 3/5 t/*: fix ordering of expected/observed argumentsMatthew DeVore, Sep 21, 2018
  22. 4/5 tests: don't swallow Git errors upstream of pipesMatthew DeVore, Sep 21, 2018
  23. 5/5 t9109: don't swallow Git errors upstream of pipesMatthew DeVore, Sep 21, 2018
  24. 0/7 Clean up tests for test_cmp arg ordering and pipe placementMatthew DeVore, Oct 3, 2018
  25. 1/7 t/README: reformat Do, Don't, Keep in mind listsMatthew DeVore, Oct 3, 2018
  26. Junio C HamanoOct 5, 2018
  27. Matthew DeVoreOct 5, 2018
  28. 2/7 Documentation: add shell guidelinesMatthew DeVore, Oct 3, 2018
  29. Junio C HamanoOct 5, 2018
  30. Matthew DeVoreOct 5, 2018
  31. 3/7 tests: standardize pipe placementMatthew DeVore, Oct 3, 2018
  32. 4/7 t/*: fix ordering of expected/observed argumentsMatthew DeVore, Oct 3, 2018
  33. 5/7 tests: don't swallow Git errors upstream of pipesMatthew DeVore, Oct 3, 2018
  34. Junio C HamanoOct 5, 2018
  35. Matthew DeVoreOct 5, 2018
  36. Matthew DeVoreOct 5, 2018
  37. 6/7 t9109: don't swallow Git errors upstream of pipesMatthew DeVore, Oct 3, 2018
  38. 7/7 tests: order arguments to git-rev-list properlyMatthew DeVore, Oct 3, 2018
  39. Matthew DeVoreOct 3, 2018
  40. Junio C HamanoOct 5, 2018
  41. 0/7 subject: Clean up tests for test_cmp arg ordering and pipe placementMatthew DeVore, Oct 5, 2018
  42. 1/7 t/README: reformat Do, Don't, Keep in mind listsMatthew DeVore, Oct 5, 2018
  43. 2/7 Documentation: add shell guidelinesMatthew DeVore, Oct 5, 2018
  44. 3/7 tests: standardize pipe placementMatthew DeVore, Oct 5, 2018
  45. 4/7 t/*: fix ordering of expected/observed argumentsMatthew DeVore, Oct 5, 2018
  46. 5/7 tests: don't swallow Git errors upstream of pipesMatthew DeVore, Oct 5, 2018
  47. 6/7 t9109: don't swallow Git errors upstream of pipesMatthew DeVore, Oct 5, 2018
  48. 7/7 tests: order arguments to git-rev-list properlyMatthew DeVore, Oct 5, 2018
  49. Junio C HamanoOct 6, 2018
  50. 1/6 CodingGuidelines: add shell piping guidelinesMatthew DeVore, Sep 17, 2018
  51. Eric SunshineSep 18, 2018
  52. Matthew DeVoreSep 19, 2018
  53. Eric SunshineSep 19, 2018
  54. Junio C HamanoSep 19, 2018
  55. Matthew DeVoreSep 19, 2018
  56. 2/6 tests: standardize pipe placementMatthew DeVore, Sep 17, 2018
  57. 3/6 t/*: fix ordering of expected/observed argumentsMatthew DeVore, Sep 17, 2018
  58. 4/6 tests: add linter check for pipe placement styleMatthew DeVore, Sep 17, 2018
  59. 5/6 tests: split up pipesMatthew DeVore, Sep 17, 2018
  60. Eric SunshineSep 18, 2018
  61. Matthew DeVoreSep 19, 2018
  62. 6/6 t9109-git-svn-props.sh: split up several pipesMatthew DeVore, Sep 17, 2018
  63. Eric SunshineSep 18, 2018
  64. Matthew DeVoreSep 19, 2018
  65. Eric SunshineSep 19, 2018
  66. Matthew DeVoreSep 19, 2018

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.