You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@apr.apache.org by "William A. Rowe, Jr." <wr...@rowe-clan.net> on 2003/01/11 00:29:05 UTC
APR Win32 Regression Status (3 failed, 2 not implemented)
What follows is the current output of testall -v on Win32, having toggled
the APR_HAVE_IPV6 to 1 (of course I'm using a modern SDK and the
developer's IPV6 driver.) Your mileage may vary with older PSDKs.
Really, this means we have exactly one real win32 failure, and two
bugs in tests themselves.
One bug concerns me, the error that "apr_stat and apr_getfileinfo differ in
protection". My working assumption is that either one, or the other, might
actually be testing a file handle's permissions (of the handle object) as
opposed to the actual object that the handle references. I hope to debug
that later tonight and close the bug for good.
I would hope the httpd team finishes regressions before blessing any
forthcoming 2.0.x release, so it would be nice if those on other platforms
might post their results as well to this list.
My recollection of OS/X is that it's nearly as clean as Win32 at this point,
and suffers from the same bogus test_localstr and test_ctime tests.
Fortunately, I'm in CST where I notice these sorts of Pacific fubars.
Finally, this doesn't imply our test suite is complete... as rbb pointed
out many times, we are testing only a fraction of the implemented
functions. I (for one) will have no problem demanding tests be committed
as new functions are deployed. Groking the otherchild logic with very
little documentation was only made possible by the example code in
the test suite.
Bill
All APR Tests:
Strings: ....
Time: ..F....F....
Versioning: ..
IP subnets: .....
MMAP: ........
User Data: ....
Table: .........
Hash: ..............
Sleep: .
Pools: ......
Formats: .....
File I/O: ....................
File Info: ..F
Pipes: .....
File duplication: ....
Directory: ..............
Random: .
DSO: .....
Other Child: .
Socket Creation: .....
Socket Options: ....N.
Process control: ...
Poll: .....................
Thread Locks: ....
Threads: .....
Getopt: ......
Path names: ...........
Users: .N.
187 tests run: 182 passed, 3 failed, 2 not implemented.
Failed tests in Time:
1) test_localstr: expected
---->
2002-08-14 12:05:36.186711 -25200 [257 Sat] DST
<----
but saw
---->
2002-08-14 14:05:36.186711 -18000 [257 Sat] DST
<----
2) test_ctime: expected
---->
Sat Sep 14 12:05:36 2002
<----
but saw
---->
Sat Sep 14 14:05:36 2002
<----
Failed tests in File Info:
1) test_stat_eq_finfo: apr_stat and apr_getfileinfo differ in protection
Not Implemented tests in Socket Options:
1) corkable: TCP isn't corkable not implemented on this platform
Not Implemented tests in Users:
1) username: Groups from apr_uid_get not implemented on this platform
Re: APR Win32 Regression Status (3 failed, 2 not implemented)
Posted by "William A. Rowe, Jr." <wr...@rowe-clan.net>.
At 05:49 PM 1/10/2003, Joe Orton wrote:
>On Fri, Jan 10, 2003 at 05:29:05PM -0600, William Rowe wrote:
>...
>> My recollection of OS/X is that it's nearly as clean as Win32 at this point,
>> and suffers from the same bogus test_localstr and test_ctime tests.
>> Fortunately, I'm in CST where I notice these sorts of Pacific fubars.
>
>Is it possible to easily change the timezone per-process on Win32 like
>it is in Unix? I guess these tests will always have to be run in *some*
>specific timezone, so maybe PST is as good as any really?
>
>(using TZ=PST8PDT ./testall seems to be a pretty portable way of getting
>these tests to pass on Unix systems)
Unfortunately, not. One of the oddities is that simply overriding the timezone
doesn't help, because of other location specifics (try Arizona as a PST (?)
locale that doesn't observe DST and also watch things choke.) Win32 timezone
changes are reflected in all processes, so your mail client might fetch up some
out-of-sequence mails. Not healthy.
So the easiest thing to do, methinks, is to fix up the resultant structure and
renormalize the value *back* to GMT based on the gmtoff member. This way
we will know that the math all adds correctly.
Yes, I have the laptop running in PST (CA) but that really isn't an answer
for a robust test suite :-)
Bill
Re: APR Win32 Regression Status (3 failed, 2 not implemented)
Posted by Joe Orton <jo...@manyfish.co.uk>.
On Fri, Jan 10, 2003 at 05:29:05PM -0600, William Rowe wrote:
...
> My recollection of OS/X is that it's nearly as clean as Win32 at this point,
> and suffers from the same bogus test_localstr and test_ctime tests.
> Fortunately, I'm in CST where I notice these sorts of Pacific fubars.
Is it possible to easily change the timezone per-process on Win32 like
it is in Unix? I guess these tests will always have to be run in *some*
specific timezone, so maybe PST is as good as any really?
(using TZ=PST8PDT ./testall seems to be a pretty portable way of getting
these tests to pass on Unix systems)
Regards,
joe
Re: APR Win32 Regression Status (3 failed, 2 not implemented)
Posted by Garrett Rooney <ro...@electricjellyfish.net>.
On Friday, January 10, 2003, at 06:29 PM, William A. Rowe, Jr. wrote:
> My recollection of OS/X is that it's nearly as clean as Win32 at this
> point,
> and suffers from the same bogus test_localstr and test_ctime tests.
> Fortunately, I'm in CST where I notice these sorts of Pacific fubars.
on my OS/X box (running version 10.2.3), i'm currently seeing all the
tests pass (other than the two that aren't implemented). of course,
this is with setting the timezone on the command line to avoid the "i'm
not in PST" failures.
-garrett
Re: APR Win32 Regression Status (3 failed, 2 not implemented)
Posted by "William A. Rowe, Jr." <wr...@rowe-clan.net>.
At 05:29 PM 1/10/2003, William A. Rowe, Jr. wrote:
>Not Implemented tests in Users:
>1) username: Groups from apr_uid_get not implemented on this platform
I should point out that this is why apr_uid_get has no business returning
apr_gid's... and should be changed when we proceed on the next release
(not breaking 0.9).
Bill