httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From William A Rowe Jr <wr...@rowe-clan.net>
Subject Re: State of trunk --enable-modules=reallyall build
Date Sat, 20 Aug 2016 18:52:32 GMT
On Fri, Aug 19, 2016 at 8:17 PM, William A Rowe Jr <wrowe@rowe-clan.net>
wrote:

> On Fri, Aug 19, 2016 at 6:14 PM, Yann Ylavic <ylavic.dev@gmail.com> wrote:
>
>>
>> How couldn't it figure out that apr_pstrcat() never returns NULL?
>> Clever compilers should really read all the docs :)
>>
>
> :)
>
>
>> Anyway, "fixed" in r1756976, thanks!
>>
>
> Thank you!
>
> My survey of --enable-modules=reallyall at the moment...
> after fixing one of my own noisy bits...
>
> httpd-2.x/modules/filters/mod_crypto.c: In function
> 'parse_pass_conf_binary':
> httpd-2.x/modules/filters/mod_crypto.c:127:9: warning: ISO C90 forbids
> mixed declarations and code [-Wdeclaration-after-statement]
>          const char *name = ap_server_root_relative(cmd->temp_pool, arg);
>          ^
>
> I expect this is on minfrin. That seems to be the remaining extent
> of all errors we caused on trunk builds, modulo ldap stuff which won't
> build against apr-2.0 and some other feature-specific stuff that
> doesn't align with the FC22 instance I'm testing on.
>
> And on FC22, with the new ar syntax, I have to deal with thousands
> of these stupid emits daily...
>
> ar: `u' modifier ignored since `D' is the default (see `U')
>
> It would be nice if any RedHat/gnu toolchain wonk who adopted this
> change would speak up about the proper fix to our configure.in.
>
> We also have breakage in the bundled FC22 bison sources, some
> clown thinks C++ comments in C sources are cute, but this appears
> to have nothing to do with httpd's build schema...
>
> util_expr_scan.c: In function 'ap_expr_yyensure_buffer_stack':
> util_expr_scan.c:2205:21: error: C++ style comments are not allowed in ISO
> C90
>    num_to_alloc = 1; // After all that talk, this was set to 1 anyways...
>                      ^
> util_expr_scan.c:2205:21: error: (this will be reported only once per
> input file)
> build/httpd2x-apr20-ossl110/build/rules.mk:207: recipe for target
> 'util_expr_scan.lo' failed
>
> That's classy.
>
> This is looking pretty good, update tomorrow on oddball t/TEST issues.
>

And as promised, 2.x + apr-2.0.0-dev with -enable-mods-shared=reallyall
against t/TEST, we immediately bust on mod_ssl_ct.so

[Sat Aug 20 17:16:18.856144 2016] [ssl_ct:error] [pid 24988:tid
140055724791936] AH02717: Directive CTSCTStorage is required
AH00014: Configuration check failed

Is anyone maintaining that bit? There's a fundamental problem any time
we can fail to start the server by simply loading a module. Some sensible
default must be present. This needs to be fixed simply to have our tests
available in this schema. Commenting out mod_ssl_ct.so...

[Sat Aug 20 18:44:00.537713 2016] [core:warn] [pid 27411:tid
140510888274048] AH00111: Config variable ${DOCROOT} is not defined

pops up on every test. Something we've changed in httpd that wasn't
reflected in the test suite, I'm guessing?

t/apache/passbrigade.t            (Wstat: 0 Tests: 114 Failed: 1)
  Failed test:  56
t/ssl/varlookup.t                 (Wstat: 0 Tests: 81 Failed: 2)
  Failed tests:  56, 58

passbrigade.t #56 is worrying. It doesn't fail consistently. I've attached a
diff of a good vs. bad error/access/ssl_access logs and t/TEST output

varlookup.t isn't really a worrying, seems to be some confusion on my
host name which couldn't be righted with -servername etc. Very clearly
an openssl cert generation issue, not a defect in httpd/mod_ssl.

# testing : SSL_SERVER_SAN_DNS_0
# expected: 'localhost'
# received: 'hub'
not ok 56
# testing : SSL_SERVER_SAN_OTHER_dnsSRV_0
# expected: '_https.localhost'
# received: '_https.hub'
not ok 58


Long story short, httpd 2.5.0-dev is looking awfully solid, as is the
test framework.

Mime
View raw message