pw32-dev

[pw32-dev] Re: zsh compiled and linked with PW32.

[Paul S. <Pau...@te...>]

Hello Earnie,

понедельник, 3 апреля 2000 г., you wrote to me:

EB> Also, BTW, the termcap-1.3 build OOB but had to
EB> add the -D_POSIX_SOURCE=1 compile switch in the Makefile.

    Well, I built it myself now. Requirement of -D_POSIX_SOURCE=1
stems from idiosyncrasy of DJGPP headers. I have no idea why Delorie
did it that way, it's violation of susv2. So, I consider that bug and
fixed it. Now it really builds OOB.

    Alike problem was with findutils. Also fixed.


EB> Some of the problems deal with lack of ioctl, SIGSTOP, SIGCONT, SIGTSTP,
EB> SIGTTOU and SIGTTIN.  There are other; but these are the most significant.

    Well, I haven't yet find time for zsh, but with bunch of other
changes I now have bash built and seemingly working (without readline,
though). Didn't run testsuite yet.

EB> Regards,
EB> Earnie.


--
Paul Sokolovsky, IT Specialist
http://www.brainbench.com/transcript.jsp?pid=11135

[pw32-dev] Re: [PW32 BUG] stat() function gives "unknown signal" error in GDB.

[Paul S. <Pau...@te...>]

Hello Earnie,

четверг, 6 апреля 2000 г., you wrote to me:

EB> ----stat-bug.c-----
EB> #include <sys/stat.h>

EB> int
EB> main ( int argc, void **argv )
EB> {
EB>   struct stat sbuf;
EB>   stat("d:/tmp/a.c", &sbuf);
EB> }
EB> ---end stat-bug.c---

EB> Attached is a bitmap of the screen during the insite debug session.  The error
EB> occurs within the call to CloseHandle.

EB> This is on a NT4 sp4.

    Well, took several minutes to investigate. Thank you, very obscure
bug! The cause is that FindFirstFile() returns value not belonging to
common kernel namespace, and it should be closed with FindClose()
instead of CloseHandle(). But see how NT is smart - it by default
ignores that 0xc000008 (invalid handle) exception, and its being
caught only under debugger. No need to say that 9x doesn't generate it
at all.

EB> Regards,
EB> Earnie.




--
Paul Sokolovsky, IT Specialist
http://www.brainbench.com/transcript.jsp?pid=11135

[pw32-dev] Re: zsh compiled and linked with PW32.

[Paul S. <Pau...@te...>]

Hello Earnie,

понедельник, 3 апреля 2000 г., you wrote to me:

EB> This is just an FYI.  I've been able to build to zsh.exe.  It doesn't quite
EB> work yet and there's a lot of work left to do; but, it's exciting to know that
EB> I got this far with your PW32.

    For me, too ;-)

EB> Also, BTW, the termcap-1.3 build OOB but had to
EB> add the -D_POSIX_SOURCE=1 compile switch in the Makefile.

    Well, but does it *work*? Nothing abot that was done from my side.
So, to make it work one of the following should be done:

- Intrinsic support for non-dumb terminal
- Relying on external means of providing non-dumb terminal

  As you guess, I'd vote for the second, if there were such method. I
know that NT has ANSI terminal driver but I have no idea whether it
installed by default and if no, where it lives at all. As for 9x,
situation even worse - ms stopped distributing ansi.sys from dos 7.
And even if such thing existed (of course it does), it's doubtful that
win32 app output will be processed by it. Darkness. So, if you've got
automagical line editing in zsh, tell me so, especially if you have
slightest idea on what make it happen ;-)

EB> Some of the problems deal with lack of ioctl, SIGSTOP, SIGCONT, SIGTSTP,
EB> SIGTTOU and SIGTTIN.  There are other; but these are the most significant.

    Well, I'll throw names in. As for functionality, SIGT* - no
chance. For SIFSTOP/SIGCONT, it's trivial, what is hard - making
stopped/continued status available via waitpid(). I have idea how to
make it on NT, but no bright solution for 9x.

EB> Regards,
EB> Earnie.


    Well, I hope to release 0.5.0 when first spring shock will dim out
;-)


--
Paul Sokolovsky, IT Specialist
http://www.brainbench.com/transcript.jsp?pid=11135