You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@httpd.apache.org by NormW <no...@gknw.net> on 2016/09/29 10:07:12 UTC

C89 (alias NetWare) tries new mod_ on the block.

G/E.
Aimed 'our' C89 compiler at mod_brotli as much for interest as anything 
technical) and get the following using release 0.5.2 of Brotli:

> Calling NWGNUmod_brotli
> CC   mod_brotli.c
> ### mwccnlm Compiler:
> #    File: mod_brotli.c
> # ---------------------
> #      22:  #include <brotli/encode.h>
> #   Error:                        ^
> #   the file 'brotli/encode.h' cannot be opened
> #   Too many errors printed, aborting program
>
> User break, cancelled...
> D:/Projects/svn/httpd-trunk/build/NWGNUtail.inc:125: recipe for target 'obj_release/mod_brotli.o' failed
> make[3]: *** [obj_release/mod_brotli.o] Error 2
> CC   D:\Projects\srcs\brotli-0.5.2/enc/backward_references.c
> ### mwccnlm Compiler:
> #    File: D:\Projects\srcs\brotli-0.5.2\enc\backward_references.c
:> #   Too many errors printed, aborting program
>
> User break, cancelled...
> D:/Projects/svn/httpd-trunk/build/NWGNUtail.inc:125: recipe for target 'obj_release/backward_references.o' failed
> make[3]: *** [obj_release/backward_references.o] Error 2
> CC   D:\Projects\srcs\brotli-0.5.2/enc/bit_cost.c
> CC   D:\Projects\srcs\brotli-0.5.2/dec/bit_reader.c
> CC   D:\Projects\srcs\brotli-0.5.2/enc/block_splitter.c
> CC   D:\Projects\srcs\brotli-0.5.2/enc/brotli_bit_stream.c
> CC   D:\Projects\srcs\brotli-0.5.2/enc/cluster.c
> CC   D:\Projects\srcs\brotli-0.5.2/enc/compress_fragment.c
> CC   D:\Projects\srcs\brotli-0.5.2/enc/compress_fragment_two_pass.c
> CC   D:\Projects\srcs\brotli-0.5.2/dec/decode.c
> CC   D:\Projects\srcs\brotli-0.5.2/common/dictionary.c
> CC   D:\Projects\srcs\brotli-0.5.2/enc/encode.c
> CC   D:\Projects\srcs\brotli-0.5.2/enc/entropy_encode.c
> CC   D:\Projects\srcs\brotli-0.5.2/enc/histogram.c
> CC   D:\Projects\srcs\brotli-0.5.2/dec/huffman.c
> CC   D:\Projects\srcs\brotli-0.5.2/enc/literal_cost.c
> CC   D:\Projects\srcs\brotli-0.5.2/enc/memory.c
> CC   D:\Projects\srcs\brotli-0.5.2/enc/metablock.c
> CC   D:\Projects\srcs\brotli-0.5.2/dec/state.c
> CC   D:\Projects\srcs\brotli-0.5.2/enc/static_dict.c
> CC   D:\Projects\srcs\brotli-0.5.2/enc/utf8_util.c
> make[3]: Target 'default' not remade because of errors.

Unclear if the ggit source has a <brotli/...> dir or this reference is 
to the brotli source as a whole).
Changing this to <enc/encode.h> and adding a -I for the brotli source 
top gets:
> Calling NWGNUmod_brotli
> GEN  obj_release/mod_brotli_cc.opt
> CC   mod_brotli.c
> ### mwccnlm Compiler:
> #    File: mod_brotli.c
> # ---------------------
> #     240:              output = BrotliEncoderTakeOutput(ctx->state, &output_len);
> #   Error:                                                                       ^
> #   illegal implicit conversion from 'int' to
> #   'const unsigned char *'
Which I assume is either a C89-ism or CW-ism..

The
>> # ----------------------------------------------------------------
>> #      30: static const float kInfinity = INFINITY;
>> #   Error:                                        ^
>> #   illegal constant expression
seems more like a C89-ism to me.

The build script and small tweak to use it follow.

Norm

NWGNUmod_brotli:
#
# Make sure all needed macro's are defined
#

#
# Get the 'head' of the build environment if necessary.  This includes 
default
# targets and paths to tools
#

ifndef EnvironmentDefined
include $(AP_WORK)/build/NWGNUhead.inc
endif

#
# These directories will be at the beginning of the include list, 
followed by
# INCDIRS
#
XINCDIRS	+= \
			$(APR)/include \
			$(APRUTIL)/include \
			$(SRC)/include \
			$(NWOS) \
			$(BROTLISRC)/common \
			$(BROTLISRC)/dec \
			$(BROTLISRC)/enc \
			$(BROTLISRC) \
			$(EOLIST)

#
# These flags will come after CFLAGS
#
XCFLAGS		+= \
			$(EOLIST)

#
# These defines will come after DEFINES
#
XDEFINES	+= \
			$(EOLIST)

#
# These flags will be added to the link.opt file
#
XLFLAGS		+= \
			$(EOLIST)

#
# These values will be appended to the correct variables based on the 
value of
# RELEASE
#
ifeq "$(RELEASE)" "debug"
XINCDIRS	+= \
			$(EOLIST)

XCFLAGS		+= \
			$(EOLIST)

XDEFINES	+= \
			$(EOLIST)

XLFLAGS		+= \
			$(EOLIST)
endif

ifeq "$(RELEASE)" "noopt"
XINCDIRS	+= \
			$(EOLIST)

XCFLAGS		+= \
			$(EOLIST)

XDEFINES	+= \
			$(EOLIST)

XLFLAGS		+= \
			$(EOLIST)
endif

ifeq "$(RELEASE)" "release"
XINCDIRS	+= \
			$(EOLIST)

XCFLAGS		+= \
			$(EOLIST)

XDEFINES	+= \
			$(EOLIST)

XLFLAGS		+= \
			$(EOLIST)
endif

#
# These are used by the link target if an NLM is being generated
# This is used by the link 'name' directive to name the nlm.  If left blank
# TARGET_nlm (see below) will be used.
#
NLM_NAME	= mod_brotli

#
# This is used by the link '-desc ' directive.
# If left blank, NLM_NAME will be used.
#
NLM_DESCRIPTION	= Apache $(VERSION_STR) Brotli Compression Module

#
# This is used by the '-threadname' directive.  If left blank,
# NLM_NAME Thread will be used.
#
NLM_THREAD_NAME	= Brotli Module

#
# If this is specified, it will override VERSION value in
# $(AP_WORK)/build/NWGNUenvironment.inc
#
NLM_VERSION	=

#
# If this is specified, it will override the default of 64K
#
NLM_STACK_SIZE	= 8192


#
# If this is specified it will be used by the link '-entry' directive
#
NLM_ENTRY_SYM	=

#
# If this is specified it will be used by the link '-exit' directive
#
NLM_EXIT_SYM	=

#
# If this is specified it will be used by the link '-check' directive
#
NLM_CHECK_SYM	=

#
# If these are specified it will be used by the link '-flags' directive
#
NLM_FLAGS	=

#
# If this is specified it will be linked in with the XDCData option in 
the def
# file instead of the default of $(NWOS)/apache.xdc.  XDCData can be 
disabled
# by setting APACHE_UNIPROC in the environment
#
XDCDATA		=

#
# If there is an NLM target, put it here
#
TARGET_nlm = \
	$(OBJDIR)/$(NLM_NAME).nlm \
	$(EOLIST)

#
# If there is an LIB target, put it here
#
TARGET_lib = \
	$(EOLIST)

#
# These are the OBJ files needed to create the NLM target above.
# Paths must all use the '/' character
#
FILES_nlm_objs = \
	$(OBJDIR)/mod_brotli.o \
	$(OBJDIR)/backward_references.o \
	$(OBJDIR)/bit_cost.o \
	$(OBJDIR)/bit_reader.o \
	$(OBJDIR)/block_splitter.o \
	$(OBJDIR)/brotli_bit_stream.o \
	$(OBJDIR)/cluster.o \
	$(OBJDIR)/compress_fragment.o \
	$(OBJDIR)/compress_fragment_two_pass.o \
	$(OBJDIR)/decode.o \
	$(OBJDIR)/dictionary.o \
	$(OBJDIR)/encode.o \
	$(OBJDIR)/entropy_encode.o \
	$(OBJDIR)/histogram.o \
	$(OBJDIR)/huffman.o \
	$(OBJDIR)/literal_cost.o \
	$(OBJDIR)/memory.o \
	$(OBJDIR)/metablock.o \
	$(OBJDIR)/state.o \
	$(OBJDIR)/static_dict.o \
	$(OBJDIR)/utf8_util.o \
	$(EOLIST)

#
# These are the LIB files needed to create the NLM target above.
# These will be added as a library command in the link.opt file.
#
FILES_nlm_libs = \
	$(PRELUDE) \
	$(EOLIST)

#
# These are the modules that the above NLM target depends on to load.
# These will be added as a module command in the link.opt file.
#
FILES_nlm_modules = \
	aprlib \
	libc \
	$(EOLIST)

#
# If the nlm has a msg file, put it's path here
#
FILE_nlm_msg =

#
# If the nlm has a hlp file put it's path here
#
FILE_nlm_hlp =

#
# If this is specified, it will override $(NWOS)\copyright.txt.
#
FILE_nlm_copyright =

#
# Any additional imports go here
#
FILES_nlm_Ximports = \
	@aprlib.imp \
	@httpd.imp \
	@libc.imp \
	$(EOLIST)

#
# Any symbols exported to here
#
FILES_nlm_exports = \
	brotli_module \
	$(EOLIST)

#
# These are the OBJ files needed to create the LIB target above.
# Paths must all use the '/' character
#
FILES_lib_objs = \
	$(EOLIST)

#
# implement targets and dependancies (leave this section alone)
#

libs :: $(OBJDIR) $(TARGET_lib)

nlms :: libs $(TARGET_nlm)

#
# Updated this target to create necessary directories and copy files to the
# correct place.  (See $(AP_WORK)/build/NWGNUhead.inc for examples)
#
install :: nlms FORCE

#
# Any specialized rules here
#
vpath %.c $(BROTLISRC)/common:$(BROTLISRC)/dec:$(BROTLISRC)/enc

#
# Include the 'tail' makefile that has targets that depend on variables 
defined
# in this makefile
#

include $(APBUILD)/NWGNUtail.inc



Index: modules/filters/NWGNUmakefile
===================================================================
--- modules/filters/NWGNUmakefile       (revision 1762749)
+++ modules/filters/NWGNUmakefile       (working copy)
@@ -154,6 +154,7 @@
  TARGET_nlm = \
         $(OBJDIR)/extfiltr.nlm \
         $(OBJDIR)/charsetl.nlm \
+        $(OBJDIR)/mod_brotli.nlm \
          $(OBJDIR)/mod_crypto.nlm \
         $(OBJDIR)/mod_data.nlm \
         $(OBJDIR)/mod_filter.nlm \


Re: C89 (alias NetWare) tries new mod_ on the block.

Posted by Gregg Smith <gl...@gknw.net>.
On 9/29/2016 4:35 AM, Evgeny Kotkov wrote:
> NormW<no...@gknw.net>  writes:
>
>> G/E.
>> Aimed 'our' C89 compiler at mod_brotli as much for interest as anything
>> technical) and get the following using release 0.5.2 of Brotli:
> Hi,
>
> Thanks for giving it a look.
>
> mod_brotli is targeted against the upcoming 1.0.x series of brotli (with a
> different include layout than 0.5.2, and with new BrotliEncoderTakeOutput()
> API that allows zero-copy processing).
>
> If you have the time and energy, could you please try this with the latest
> development snapshot from https://github.com/google/brotli ?
> This error happens because 0.5.2 doesn't have the 
> BrotliEncoderTakeOutput()
> API.  As it's undefined, the compiler assumes it returning an int, and
> that results in the error.

Hi,

Just FYI, I gave this module a try building with Visual Studio 2008 
(C89) against Apache 2.4.23 and it works quite nice. I did pick up your 
note about the upcoming  brotli 1.0.0 so I used the master.

Server version: Apache/2.4.23 (Win32)
Server built:   Sep 24 2016 00:52:57

HTTP/2.0 200 OK
Date: Thu, 29 Sep 2016 17:11:20 GMT
Server: Apache/2.4.23 (Win32) OpenSSL/1.0.2i
Last-Modified: Thu, 29 Sep 2016 17:11:14 GMT
Etag: "1234-53da890a8e938-br"
Accept-Ranges: bytes
Vary: Accept-Encoding
Content-Encoding: br
Content-Length: 1746
Content-Type: text/html
X-Firefox-Spdy: h2

Thanks for the new module! I look forward to the additions you mentioned 
when introducing it.

Cheers,
Gregg


Re: C89 (alias NetWare) tries new mod_ on the block.

Posted by Evgeny Kotkov <ev...@visualsvn.com>.
NormW <no...@gknw.net> writes:

> G/E.
> Aimed 'our' C89 compiler at mod_brotli as much for interest as anything
> technical) and get the following using release 0.5.2 of Brotli:
>
>> Calling NWGNUmod_brotli
>> CC   mod_brotli.c
>> ### mwccnlm Compiler:
>> #    File: mod_brotli.c
>> # ---------------------
>> #      22:  #include <brotli/encode.h>
>> #   Error:                        ^
>> #   the file 'brotli/encode.h' cannot be opened
>> #   Too many errors printed, aborting program

Hi,

Thanks for giving it a look.

mod_brotli is targeted against the upcoming 1.0.x series of brotli (with a
different include layout than 0.5.2, and with new BrotliEncoderTakeOutput()
API that allows zero-copy processing).

If you have the time and energy, could you please try this with the latest
development snapshot from https://github.com/google/brotli ?

>> Calling NWGNUmod_brotli
>> GEN  obj_release/mod_brotli_cc.opt
>> CC   mod_brotli.c
>> ### mwccnlm Compiler:
>> #    File: mod_brotli.c
>> # ---------------------
>> #     240:              output = BrotliEncoderTakeOutput(ctx->state,
>> &output_len);
>> #   Error:
>> ^
>> #   illegal implicit conversion from 'int' to
>> #   'const unsigned char *'
>
> Which I assume is either a C89-ism or CW-ism..

This error happens because 0.5.2 doesn't have the BrotliEncoderTakeOutput()
API.  As it's undefined, the compiler assumes it returning an int, and
that results in the error.

> The
>>>
>>> # ----------------------------------------------------------------
>>> #      30: static const float kInfinity = INFINITY;
>>> #   Error:                                        ^
>>> #   illegal constant expression
>
> seems more like a C89-ism to me.

Unfortunately, this comes from the brotli itself, and would probably require
a fix in the upstream:

  https://github.com/google/brotli/blob/8148001/enc/backward_references.c#L30

Not too sure on what is the issue here.  (How does the #define INFINITY
look like in your environment?  Perhaps, it's specifically designed to
prohibit assignment?).  One workaround to get past this would be to use
the #else part of this code.


Regards,
Evgeny Kotkov