Thread: Build broken with nmake since ea74e47
Hi, Since commit ea74e47, the build with nmake is broken: NMAKE : fatal error U1073: don't know how to make 'exe\connect-test.exe' I have yet to diagnose what is wrong here, but a guess is that win.mak is missing a trick. Thanks, -- Michael
On Thu, Jun 30, 2016 at 12:45 AM, Michael Paquier <michael.paquier@gmail.com> wrote: > Hi, > > Since commit ea74e47, the build with nmake is broken: > NMAKE : fatal error U1073: don't know how to make 'exe\connect-test.exe' > I have yet to diagnose what is wrong here, but a guess is that win.mak > is missing a trick. After spending some time on that, I can reproduce the problem, but I have little idea how to properly link each exe and src file with common.obj at the same time using this nmake Makefile. ea74e47 has really introduced a broken concept in win.mak, and I am sure that this has not been tested. Could it be possible to get back to the prior behavior for just nmake? For example the patch attached does the job for me again. At the same time, I have noticed a compilation error in large-object-data-at-exec-test.c regarding a type declaration in the middle of a code block, basically that: - int error = SQL_SUCCESS != rc && SQL_SUCCESS_WITH_INFO != rc && SQL_NEED_DATA != rc; + error = SQL_SUCCESS != rc && + SQL_SUCCESS_WITH_INFO != rc && + SQL_NEED_DATA != rc; -- Michael
Attachment
Hi Michael, On 2016/07/01 13:27, Michael Paquier wrote: > On Thu, Jun 30, 2016 at 12:45 AM, Michael Paquier > <michael.paquier@gmail.com> wrote: >> Hi, >> >> Since commit ea74e47, the build with nmake is broken: >> NMAKE : fatal error U1073: don't know how to make 'exe\connect-test.exe' >> I have yet to diagnose what is wrong here, but a guess is that win.mak >> is missing a trick. > After spending some time on that, I can reproduce the problem, but I > have little idea how to properly link each exe and src file with > common.obj at the same time using this nmake Makefile. ea74e47 has > really introduced a broken concept in win.mak, and I am sure that this > has not been tested. Could it be possible to get back to the prior > behavior for just nmake? For example the patch attached does the job > for me again. Oops my mistake. The commit was irrelevant to Windows.. I would put back the change. Thanks. > > At the same time, I have noticed a compilation error in > large-object-data-at-exec-test.c regarding a type declaration in the > middle of a code block, basically that: > - int error = SQL_SUCCESS != rc && SQL_SUCCESS_WITH_INFO != rc > && SQL_NEED_DATA != rc; > + error = SQL_SUCCESS != rc && > + SQL_SUCCESS_WITH_INFO != rc && > + SQL_NEED_DATA != rc; OK I would fix the compilation error. Thanks. Hiroshi Inoue
Hi Michael, On 2016/07/02 20:18, Inoue, Hiroshi wrote: > Hi Michael, > > On 2016/07/01 13:27, Michael Paquier wrote: >> On Thu, Jun 30, 2016 at 12:45 AM, Michael Paquier >> <michael.paquier@gmail.com> wrote: >>> Hi, >>> >>> Since commit ea74e47, the build with nmake is broken: >>> NMAKE : fatal error U1073: don't know how to make >>> 'exe\connect-test.exe' >>> I have yet to diagnose what is wrong here, but a guess is that win.mak >>> is missing a trick. >> After spending some time on that, I can reproduce the problem, but I >> have little idea how to properly link each exe and src file with >> common.obj at the same time using this nmake Makefile. ea74e47 has >> really introduced a broken concept in win.mak, and I am sure that this >> has not been tested. Could it be possible to get back to the prior >> behavior for just nmake? For example the patch attached does the job >> for me again. > > Oops my mistake. > The commit was irrelevant to Windows.. > I would put back the change. > Thanks. After looking at regression test on Windows, I'm inclined to go forward rather than putting back win.mak, i.e. build obj and exe in exe directory. I don't want to put them in src directory because I specified src directory as EXTRA_DIST. I would also use fc command instead of diff command on Windows. regards, Hiroshi Inoue
On Wed, Jul 6, 2016 at 9:51 PM, Inoue, Hiroshi <h-inoue@dream.email.ne.jp> wrote: > After looking at regression test on Windows, I'm inclined to go forward > rather > than putting back win.mak, i.e. build obj and exe in exe directory. I don't > want > to put them in src directory because I specified src directory as > EXTRA_DIST. > I would also use fc command instead of diff command on Windows. No objections from here to do that. I'll just adapt my builds to the situation: I have for example a couple of things where the path to the compiled binaries of the regression tests is hardcoded because being tarball'd and sent elsewhere, but that would be a trivial change. -- Michael
Hi Michael, On 2016/07/07 15:58, Michael Paquier wrote: > On Wed, Jul 6, 2016 at 9:51 PM, Inoue, Hiroshi > <h-inoue@dream.email.ne.jp> wrote: >> After looking at regression test on Windows, I'm inclined to go forward >> rather >> than putting back win.mak, i.e. build obj and exe in exe directory. I don't >> want >> to put them in src directory because I specified src directory as >> EXTRA_DIST. >> I would also use fc command instead of diff command on Windows. > No objections from here to do that. I'll just adapt my builds to the > situation: I have for example a couple of things where the path to the > compiled binaries of the regression tests is hardcoded because being > tarball'd and sent elsewhere, but that would be a trivial change. I pushed the changes about regression test on Windows to git. Please try. 7 tests fail here. regards, Hiroshi Inoue