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

Re: [PATCH 1/3] t5004: test ZIP archives with many entries

From
Eric Sunshine <sunshine@sunshineco.com>
Date
Aug 23, 2015, 17:45 UTC
Message-ID
<CAPig+cSNSfpt7gOLvz7P4oDrNF5fTQ38v1pfncJU3h7a6FjMyQ@mail.gmail.com>
In-Reply-To
<trinity-6e67d416-0a61-4e73-9779-63519dd83fdb-1440322151491@3capp-webde-bs47>
On Sun, Aug 23, 2015 at 5:29 AM, "René Scharfe" <l.s.r@web.de> wrote:
Show 35 quoted lines
> Am 23.08.2015 um 07:54 schrieb Eric Sunshine:
>> On Sat, Aug 22, 2015 at 3:06 PM, René Scharfe <l.s.r@web.de> wrote:
>>> +test_lazy_prereq ZIPINFO '
>>> +       n=$("$ZIPINFO" "$TEST_DIRECTORY"/t5004/empty.zip | sed -n "2s/.* //p")
>>> +       test "x$n" = "x0"
>>> +'
>>
>> Unfortunately, this sed expression isn't portable due to dissimilar
>> output of various zipinfo implementations. On Linux, the output of
>> zipinfo is:
>>
>>      $ zipinfo t/t5004/empty.zip
>>      Archive:  t/t5004/empty.zip
>>      Zip file size: 62 bytes, number of entries: 0
>>      Empty zipfile.
>>      $
>>
>> however, on Mac OS X:
>>
>>      $ zipinfo t/t5004/empty.zip
>>      Archive:  t/t5004/empty.zip   62 bytes   0 files
>>      Empty zipfile.
>>      $
>>
>> and on FreeBSD, the zipinfo command seems to have been removed
>> altogether in favor of "unzip -Z" (emulate zipinfo).
>
> I suspected that zipinfo's output might be formatted differently on
> different platforms and tried to guard against it by checking for the
> number zero there. Git's ZIP file creation is platform independent
> (modulo bugs), so having a test run at least somewhere should
> suffice. In theory.
>
> We could add support for the one-line-summary variant on OS X easily,
> though.

Probably, although it's looking like testing on Mac OS X won't be fruitful (see below).

Show 28 quoted lines
>> One might hope that "unzip -Z" would be a reasonable replacement for
>> zipinfo, however, it is apparently only partially implemented on
>> FreeBSD, and requires that -1 be passed, as well. Even with "unzip -Z
>> -1", there are issues. The output on Linux and Mac OS X is:
>>
>>      $ unzip -Z -1 t/t5004/empty.zip
>>      Empty zipfile.
>>      $
>>
>> but FreeBSD differs:
>>
>>      $ unzip -Z -1 t/t5004/empty.zip
>>      $
>>
>> With a non-empty zip file, the output is identical on all platforms:
>>
>>      $ unzip -Z -1 twofiles.zip
>>      file1
>>      file2
>>      $
>>
>> So, if you combine that with "wc -l" or test_line_count, you may have
>> a portable and reliable entry counter.
>
> Counting all entries is slow, and more importantly it's not what we
> want. In this test we need the number of entries recorded in the ZIP
> directory, not the actual number of entries found by scanning the
> archive, or the directory.
Ah, right. The commit message did state this clearly enough...
Show 33 quoted lines
> On Linux "unzip -Z -1 many.zip | wc -l" reports 65792 even before
> adding ZIP64 support; only without -1 we get the interesting numbers
> (specifically with "unzip -Z many.zip | sed -n '2p;$p'"):
>
>     Zip file size: 6841366 bytes, number of entries: 256
>     65792 files, 0 bytes uncompressed, 0 bytes compressed: 0.0%
>
>> With these three patches applied, Mac OS X has trouble with 'many.zip':
>>
>>      $ unzip -Z -1 many.zip
>>      warning [many.zip]:  76 extra bytes at beginning or within zipfile
>>        (attempting to process anyway)
>>      error [many.zip]:  reported length of central directory is
>>        -76 bytes too long (Atari STZip zipfile?  J.H.Holm ZIPSPLIT 1.1
>>        zipfile?).  Compensating...
>>      00/
>>      00/00
>>      ...
>>      ff/ff
>>      error: expected central file header signature not found (file
>>        #65793). (please check that you have transferred or created the
>>        zipfile in the appropriate BINARY mode and that you have compiled
>>        UnZip properly)
>>
>> And FreeBSD doesn't like it either:
>>
>>      $ unzip -Z -1 many.zip
>>      unzip: Invalid central directory signature
>>      $
>>
>
> Looks like they don't support ZIP64. Or I got some of the fields wrong
> after all.

A >65536 file zip created on Mac OS X with Mac's "zip" command given to "unzip" or "zipinfo" results in exactly the same warnings/errors as above (including the bit about "76 extra bytes" and "-76 bytes too long"), so it doesn't seem to be a problem with your implementation.

Show 8 quoted lines
> https://en.wikipedia.org/wiki/Zip_%28file_format%29#ZIP64 says: "OS X
> Yosemite does support the creation of ZIP64 archives, but does not
> support unzipping these archives using the shipped unzip command-line
> utility or graphical Archive Utility.[citation needed]".
>
> How does unzip react to a ZIP file with more than 65535 entries that
> was created natively on these platforms? And what does zipinfo (a real
> one, without -1) report at the top for such files?

On Mac OS X, unzip does extract all the files (although complains as noted above). zipinfo caps out at reporting 65535 for the number of files (although it lists them all fine). With the warnings/errors filtered out for clarity:

    $ zipinfo biggy.zip
    Archive:  biggy.zip   9642874 bytes   65535 files
    ...
Previous: Eric SunshineNext: Junio C Hamano
Message 10 of 18 in “bug: git-archive does not use the zip64 extension for archives with more than 16k entries”
  1. Johannes SchauerAug 11, 2015
  2. René ScharfeAug 12, 2015
  3. Johannes SchauerAug 13, 2015
  4. 1/3 t5004: test ZIP archives with many entriesRené Scharfe, Aug 22, 2015
  5. Eric SunshineAug 23, 2015
  6. René ScharfeAug 23, 2015
  7. Eric Sunshine mail delivery failureRené Scharfe, Aug 23, 2015
  8. Johannes LöthbergAug 23, 2015
  9. Eric SunshineAug 23, 2015
  10. Eric SunshineAug 23, 2015
  11. Junio C HamanoAug 28, 2015
  12. Junio C HamanoAug 28, 2015
  13. Eric SunshineAug 28, 2015
  14. 2/3 archive-zip: use a local variable to store the creator versionRené Scharfe, Aug 22, 2015
  15. 3/3 archive-zip: support more than 65535 entriesRené Scharfe, Aug 22, 2015
  16. Duy NguyenAug 15, 2015
  17. Eric SunshineAug 23, 2015
  18. Eric SunshineAug 23, 2015

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.