On Sun, Jan 27, 2013 at 3:12 AM, Kornel Benko <kor...@lyx.org> wrote:
> Am Sonntag, 27. Januar 2013 um 02:10:58, schrieb Scott Kostyshak
> <skost...@lyx.org>
>
> ....
>
>> >>
>
>> >> Should be easy.
>
>> >
>
>> > Ah yes you're right. You are convincing me more and more that
>
>> > out-of-source builds have many advantages and no disadvantages. So all
>
>> > is needed is just to copy the .lyx file there.
>
>> >
>
>> > Scott
>
>>
>
>> Please see the attached patches. Everything works fine under CMake,
>
>> except that another mystery (for me, at least) has appeared. The new
>
>> autotest only works for me if I revert the XVKBD_HACKED commit
>
>> (dae22baf). If that commit is left in, everything works fine except
>
>> that \[Return] does not get rid of the warning that is given on
>
>> opening the .lyx file. If you quickly press return manually, the rest
>
>> of the test works fine. This same behavior happens under autotools
>
>> with XVKBD_HACKED set. (Note that I haven't figured out yet how to run
>
>> the autotools tests without setting XVKBD_HACKED so I can't test
>
>> that.)
>
>>
>
>> Why would there be this difference depending on XVKBD_HACKED? The most
>
>> surprising thing is that my system version of xvkbd was installed from
>
>> the version in LyX (with make install) so I don't think it could be
>
>> differences in the versions.
>
>
>
> This surprises me. Checking the created package, there is no xvkbd
> contained.
>
> Also, we have no 'instal()' statement for xvkbd in the cmake-files.
I'm not sure what to do then. bug-8523-in.txt only works for me if
XVKBD_HACKED is not hard-coded.
>
>
>
>> Note that the attached autotest should currently fail because #8523 is
>
>> not fixed yet.
>
>>
>
>> Any ideas?
>
>>
>
>> Kornel, do you have any comments on the CMake patch?
>
>
>
> Yes.
>
> single-test.cmake:
>
>
>
> 1.) What if KEYTEST_INFILE = "abcd-in.txtabcd-in.sh"?
>
> Then _jj is "abcdabcd-in.sh", but it should be "abcd-in.txtabcd"
>
> Therefore (notice the '$')
>
> string(REGEX REPLACE "-in\\.(txt|sh)$" "" _jj ${KEYTEST_INFILE})
Good point. I will change this.
> 2.) We do not need BUILD_DIR, why not use "${WORKDIR}/.."?
This is much nicer. I did not know you could do that in CMake. For
example, I don't think you can use "~" to refer to /home/user/. Ah,
but that's because ~ is Unix-specific I guess.
> 3.) I wonder, why configure_file() is called that way.
>
> configure_file("${AUTOTEST_ROOT}/${_jj}.lyx" ${BUILD_DIR})
>
>
>
> it should at least be
>
> configure_file("${AUTOTEST_ROOT}/${_jj}.lyx" "${BUILD_DIR}/${_jj}.lyx")
>
>
>
> but even that seams not needed yet, since there are no '@xyzzy@' strings in
> bug-8523.lyx
>
>
>
> If you want to copy so
>
> execute_program(COMMAND ${CMAKE_COMMAND} -E copy
> "${AUTOTEST_ROOT}/${_jj}.lyx" "${BUILD_DIR}/${_jj}.lyx")
>
> or
>
> configure_file("${AUTOTEST_ROOT}/${_jj}.lyx" "${BUILD_DIR}/${_jj}.lyx"
> COPYONLY)
>
>
>
> should be enough.
OK, I will change this. I am in the habit of writing copies of the
form "cp filename folder1/folder2/" because (1) this prevents type
errors such as "cp filename folder1/folder2/fielname" and (2) the
reader of the code doesn't have to check whether the two filenames are
exactly the same. But I guess with this form there are concerns that
filename could overwrite folder2?
> bug-8523-in.txt:
>
>
>
> I would change the file, there is also no need to remove, on next test it
> will be overwritten anyway
>
> and we can check the result too.
>
>
>
> TestBegin ../bug-8523.lyx > lyx-log.txt 2>&1
>
> KK: \[Home]
>
> KK: some text
>
> KK: \Ax
>
> KK: buffer-save\[Return]
>
> TestEnd
>
>
I don't understand. First, on opening bug-8523.lyx you should
immediately see a warning which must be addressed. This is where the
\[Return] works if XVKBD_HACKED is not automatically set. Also, I
don't see a way around saving to another file because that is what
triggers the bug.
Thanks a lot for reviewing the patches,
Scott
