[Scons-users] 回复: 回复: 回复: 回复: 回复: Ninja compilation fails after using include path file instead of "-I" option

Daniel Moody dmoody256 at gmail.com
Fri Apr 15 11:50:27 EDT 2022


hello liruncong,

If you want to add multiple @file dependencies via the scanner do them all
in the same scanner. Like I said you can only apply one scanner, so in this
case you are overwriting each previous scanner. The scanner should return a
list of files in which to set dependencies for the current node.

The path function is not used in this case, and I just left the CPPPATH
from the default SourceFileScanner I copied the scanner code from. It does
have to be something besides None I believe though, or scons won't scan.
The correct way this should be used in this case is you should place
the @file in a known list of directories or just a single directory, and
then pass a list of directories or a list containing the single directory,
and the scanner should find the files by extension in the list of
directories. This is the fundamental working of scanners if you want to
read more: https://scons.org/doc/production/HTML/scons-user/ch20.html,
however you could also use a dummy list if there is no dynamic component
here.

Here is an updated example based on the new info you provided:

incPathsFile = env.Textfile(
    "build/__cpp_path.txt",
    cppPathStr
)

defineOptsFile = env.Textfile(
    "build/__define_options.txt",
    "\n".join(rtconfig.rtconfigObj.DefineOptions)
)

linkObjsFile = env.Textfile(
    "build/__link_objects.txt",
    "\n".join(objsList)
)

def _add_scanner(builder):
    def new_scanner(node, env, path):
        return incPathsFile + defineOptsFile + linkObjsFile

    builder.builder.target_scanner = SCons.Scanner.Scanner(
        function=new_scanner,
        path_function=[],
    )
for object_builder in SCons.Tool.createObjBuilders(env):
    _add_scanner(object_builder)

On Fri, Apr 15, 2022 at 9:58 AM liruncong2018 <liruncong2018 at qq.com> wrote:

> Can you give me more information? What source file? Does the compilation
> fail the first time? Does it recompile successfully with only scons
> building? Can you push your latest code so I can see it in your repo?
> -- I don't know how to reproduce this problem in the git test project, I
> will try to see if it can be reproduced.
> Also each builder can only hold one target or source scanner, but you can
> chain the scanner by calling other scanners from within the single scanner.
> Are you assigning multiple scanners to the same builder?
> -- The following are some scanners, I don't quite understand the
> path_function, especially the second parameter of the last scanner
> (linkObjsFile).
> ————————————————————————————————————————
>     incPathsFile = env.Textfile(
>         "build/__cpp_path.txt",
>         cppPathStr
>     )
>     def _add_scanner(builder):
>         def new_scanner(node, env, path):
>             return incPathsFile
>
>         builder.builder.target_scanner = SCons.Scanner.Scanner(
>             function=new_scanner,
>             path_function=SCons.Script.FindPathDirs('CPPPATH'),
>         )
>     for object_builder in SCons.Tool.createObjBuilders(env):
>         _add_scanner(object_builder)
>
>     defineOptsFile = env.Textfile(
>         "build/__define_options.txt",
>         "\n".join(rtconfig.rtconfigObj.DefineOptions)
>     )
>
>     def _add_scanner(builder):
>         def new_scanner(node, env, path):
>             return defineOptsFile
>
>         builder.builder.target_scanner = SCons.Scanner.Scanner(
>             function=new_scanner,
>             path_function=SCons.Script.FindPathDirs('CPPPATH'),
>         )
>     for object_builder in SCons.Tool.createObjBuilders(env):
>         _add_scanner(object_builder)
>
>     linkObjsFile = env.Textfile(
>         "build/__link_objects.txt",
>         "\n".join(objsList)
>     )
>
>     def _add_scanner(builder):
>         def new_scanner(node, env, path):
>             return linkObjsFile
>
>         builder.builder.target_scanner = SCons.Scanner.Scanner(
>             function=new_scanner,
>             path_function=SCons.Script.FindPathDirs('LIBPATH'),
>         )
>     for object_builder in SCons.Tool.createObjBuilders(env):
>         _add_scanner(object_builder)
> ————————————————————————————————————————————————
> ------------------ 原始邮件 ------------------
> *发件人:* "Daniel Moody" <dmoody256 at gmail.com>;
> *发送时间:* 2022年4月14日(星期四) 晚上11:36
> *收件人:* "liruncong2018"<liruncong2018 at qq.com>;
> *抄送:* "SCons users mailing list"<scons-users at scons.org>;
> *主题:* Re: [Scons-users] 回复: 回复: 回复: 回复: 回复: Ninja compilation fails after
> using include path file instead of "-I" option
>
> Can you give me more information? What source file? Does the compilation
> fail the first time? Does it recompile successfully with only scons
> building? Can you push your latest code so I can see it in your repo?
>
>  Also each builder can only hold one target or source scanner, but you can
> chain the scanner by calling other scanners from within the single scanner.
> Are you assigning multiple scanners to the same builder?
>
>
>
> On Thu, Apr 14, 2022 at 10:17 AM liruncong2018 <liruncong2018 at qq.com>
> wrote:
>
>> Hello Daniel Moody,
>> Thanks a lot for the example you gave. According to your example, I added
>> 4 scanners, which can indeed be compiled and linked.
>> But when I edit the source file, I deliberately put "a" at the end of the
>> .cpp file and expect the compilation to fail. But when I compile again,
>> nothing compiles.
>> How to diagnose this problem?
>>
>> ------------------ 原始邮件 ------------------
>> *发件人:* "Daniel Moody" <dmoody256 at gmail.com>;
>> *发送时间:* 2022年4月14日(星期四) 凌晨2:27
>> *收件人:* "liruncong2018"<liruncong2018 at qq.com>;
>> *抄送:* "SCons users mailing list"<scons-users at scons.org>;
>> *主题:* Re: [Scons-users] 回复: 回复: 回复: 回复: 回复: Ninja compilation fails
>> after using include path file instead of "-I" option
>>
>> I added an example how to hook up the dependencies:
>>
>> https://github.com/liruncong/NinJaTest/pull/1/files
>>
>> On Wed, Apr 13, 2022 at 10:01 AM liruncong2018 <liruncong2018 at qq.com>
>> wrote:
>>
>>> Hi,
>>> I generate some option files via env.Textfile ("build/__cpp_path.txt",
>>> "build/__define_options.txt", "build/__link_objects.txt",
>>> "build/__exe_cpp_path.txt") and make objs depend on corresponding files
>>> ,for example:
>>> env.Depends(objs, ["build/__cpp_path.txt", "build/__define_options.txt"
>>> ])
>>> env.Depends(target, "build/__link_objects.txt")
>>> But after enabling ninja, these manually added dependencies are ignored,
>>> which causes ninja to fail to compile.
>>> How to deal with this?
>>>
>>> ------------------ 原始邮件 ------------------
>>> *发件人:* "Daniel Moody" <dmoody256 at gmail.com>;
>>> *发送时间:* 2022年4月13日(星期三) 凌晨0:02
>>> *收件人:* "SCons users mailing list"<scons-users at scons.org>;
>>> *抄送:* "liruncong2018"<liruncong2018 at qq.com>;
>>> *主题:* Re: [Scons-users] 回复: 回复: 回复: 回复: 回复: Ninja compilation fails
>>> after using include path file instead of "-I" option
>>>
>>> Hello, that PR is meant to support shared @file with ninja, not intended
>>> for any improvements to TEMPFILE. Please test without TEMPFILE.
>>>
>>> On Tue, Apr 12, 2022 at 10:52 AM liruncong2018 via Scons-users <
>>> scons-users at scons.org> wrote:
>>>
>>>> Hi
>>>> I tested "https://github.com/SCons/scons/pull/4133", after enabling
>>>> ningja, except for the link failure, .c/.cpp/.S are successfully compiled
>>>> into .o, the total time is 6:48 .
>>>> This time is much greater than the previous time of 3:45 when ninja was
>>>> not enabled and the option file was shared using @file.
>>>> After using TEMPFILE instead of @file, the compilation time is
>>>> basically doubled whether ninja is enabled or not.
>>>> The TEMPFILE function should be improved to use shared files to prevent
>>>> the creation of a large number of temporary files.
>>>>
>>>> ------------------ 原始邮件 ------------------
>>>> *发件人:* "SCons users mailing list" <dmoody256 at gmail.com>;
>>>> *发送时间:* 2022年4月11日(星期一) 晚上11:43
>>>> *收件人:* "SCons users mailing list"<scons-users at scons.org>;
>>>> *主题:* Re: [Scons-users] 回复: 回复: 回复: 回复: Ninja compilation fails after
>>>> using include path file instead of "-I" option
>>>>
>>>> I have a potential fix to the ninja tool so that it can decide to use
>>>> response files or not. This should allow you to use your own shared
>>>> response file.
>>>>
>>>> https://github.com/SCons/scons/pull/4133
>>>>
>>>> On Mon, Apr 11, 2022 at 10:01 AM Mats Wichmann <mats at wichmann.us>
>>>> wrote:
>>>>
>>>>>
>>>>> On 4/11/22 08:57, liruncong2018 via Scons-users wrote:
>>>>> > Hi,
>>>>> > Even using msys2 does not solve all problems. The 32K limit solves my
>>>>> > compilation problems (.cpp -> .o) but not linking (.o -> target)
>>>>> because
>>>>> > the link command line has around 137K (about 2500 .o files). So
>>>>> > compilers such as gcc/clang/armclang will support "@file".
>>>>> > Currently, scons needs to set up Depends when using @file. This
>>>>> > dependency suggestion is still directly supported by scons, and
>>>>> users do
>>>>> > not need to write dependencies. It should also be supported when
>>>>> ninja
>>>>> > is enabled.
>>>>> > When LINKCOM uses "TEMPFILE", I get a link failure because
>>>>> armlink.exe
>>>>> > checks if --cpu is specified on the command line, so it is not
>>>>> > reasonable for "TEMPFILE" to put all options to a temporary file.
>>>>>
>>>>> Right... thus my suggestion - well, let's call it "musing" rather than
>>>>> "suggestion" because this might not work at all:
>>>>>
>>>>> > For ninja it currently looks like this:
>>>>> >
>>>>> > env["LINKCOM"] = '${TEMPFILE("$LINK $LINKFLAGS /OUT:$TARGET.windows
>>>>> > $_LIBDIRFLAGS $_LIBFLAGS $_PDB $SOURCES.windows", "$LINKCOMSTR")}'
>>>>> >
>>>>> > env["SHLINKCOM"] = '${TEMPFILE("$SHLINK $SHLINKFLAGS $_SHLINK_TARGETS
>>>>> > $_LIBDIRFLAGS $_LIBFLAGS $_PDB $_SHLINK_SOURCES", "$SHLINKCOMSTR")}'
>>>>> >
>>>>> > So I'm curious if moving $LINK $LINKFLAGS to the left of ${TEMPFILE
>>>>> > would help...
>>>>> _______________________________________________
>>>>> Scons-users mailing list
>>>>> Scons-users at scons.org
>>>>> https://pairlist4.pair.net/mailman/listinfo/scons-users
>>>>>
>>>> _______________________________________________
>>>> Scons-users mailing list
>>>> Scons-users at scons.org
>>>> https://pairlist4.pair.net/mailman/listinfo/scons-users
>>>>
>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://pairlist4.pair.net/pipermail/scons-users/attachments/20220415/5aa037c6/attachment-0001.htm>


More information about the Scons-users mailing list