lzip-bug
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: ChangeLog encoding; testsuite file/s


From: Antonio Diaz Diaz
Subject: Re: ChangeLog encoding; testsuite file/s
Date: Tue, 25 Jan 2022 16:57:26 +0100
User-agent: Mozilla/5.0 (X11; U; Linux i586; en-US; rv:1.9.1.19) Gecko/20110420 SeaMonkey/2.0.14

Adam Tuja wrote:
Reason I brought it is, there is (almost) always way to do something better.
And also observation, that `test.txt/.lz' is "doubled" `COPYING' from source
directory. Therefore it can be generated.

Generating test files from files outside of the testsuite directory is out of the question because it causes trouble to downstream packagers.

Moreover, 'test.txt' is NOT a "doubled" 'COPYING'. The second "copy" has CRLF line endings not possible to reliably generate, and required to test lzip on non-posix systems.

COPYING must be in the tarball for legal reasons. Therefore including in the same tarball a test.txt derived from it just increases the size of the compressed tarball by about 1 KB.

$ cat COPYING | lzip -9v > /dev/null
  (stdin):  2.770:1, 36.10% ratio, 63.90% saved, 18025 in, 6507 out.
$ cat COPYING testsuite/test.txt | lzip -9v > /dev/null
  (stdin):  7.305:1, 13.69% ratio, 86.31% saved, 54413 in, 7449 out.


Same with `test_em.lz', something along those lines:
...
printf "">  0

'printf' does not work reliably on all systems were lzip can be compiled. See for example the warning at line 374 of check.sh:

printf "\nwarning: skipping header test: 'printf' does not work on your system."

Your change also ignores completely the testing of some lzip features. For example, it would break the testing of volume creation (check.sh line 283) and the truncation test (check.sh line 391). It would also break the testsuites of other programs like plzip and lziprecover.


The last, as you mentioned gzip - all tests there are generated, and size is due
to the sources, not tests.

Gzip-1.11 contains at least one non-generated test file (tests/hufts-segv.gz) because not everyting can be generated, and even what can be generated may not be wise to do so if one wants to be sure that the test file is identical on all systems.

BTW, the testsuite of gzip is twice as large as that of lzip:

$ cat lzip-1.23-rc1/testsuite/* | wc -c
74949
$ cat gzip-1.11/tests/* | wc -c
147064


Regards,
Antonio.



reply via email to

[Prev in Thread] Current Thread [Next in Thread]