Has anyone tried to run TCL 8.x tests on RTP?
It compiles relatively clean, except that configure does not detect that
RTP has fd_set (probably because it’s typedef, rather than #define) and
then fd_mask is undefined (should be int).
But there are so many problems with tests that I’m afraid something is
severely broken. Aside of numerous failing tests (strings, parsing,
float precision, some others) there are tests which can’t be executed at
all (basic, event, io, http, httpold). They simply never return, either
spinning in READY or stuck in SIGWAITINFO forever. And if I’m running
big TCP stack instead of tiny then some tests hang the system
altogether.
The stack test fails with SIGSEGV.
I’m not at all familiar with TCL to look into tests, unfortunately. Any
volunteers to dig the issue?
You could post the test results to comp.lang.tcl, they’re a pretty helpful
bunch.
FWIW I’m using tcl 8.4a2 on QNX 4.25.
-Bruce.
Igor Kovalenko wrote:
Has anyone tried to run TCL 8.x tests on RTP?
It compiles relatively clean, except that configure does not detect that
RTP has fd_set (probably because it’s typedef, rather than #define) and
then fd_mask is undefined (should be int).
But there are so many problems with tests that I’m afraid something is
severely broken. Aside of numerous failing tests (strings, parsing,
float precision, some others) there are tests which can’t be executed at
all (basic, event, io, http, httpold). They simply never return, either
spinning in READY or stuck in SIGWAITINFO forever. And if I’m running
big TCP stack instead of tiny then some tests hang the system
altogether.
The stack test fails with SIGSEGV.
I’m not at all familiar with TCL to look into tests, unfortunately. Any
volunteers to dig the issue?
Has anyone tried to run TCL 8.x tests on RTP?
It compiles relatively clean, except that configure does not detect that
RTP has fd_set (probably because it’s typedef, rather than #define) and
then fd_mask is undefined (should be int).
But there are so many problems with tests that I’m afraid something is
severely broken. Aside of numerous failing tests (strings, parsing,
float precision, some others) there are tests which can’t be executed at
all (basic, event, io, http, httpold). They simply never return, either
spinning in READY or stuck in SIGWAITINFO forever. And if I’m running
big TCP stack instead of tiny then some tests hang the system
altogether.
The stack test fails with SIGSEGV.
I’m not at all familiar with TCL to look into tests, unfortunately. Any
volunteers to dig the issue?
At one point in time we had someone in-house who had done the full
TCL/Tk port and had managed (if I remember correctly) to get it
to pass through all of it’s tests cleanly. It would be my hope
that our diffs/mods were sent back for inclusion but I’m only
guessing. Colin would know more … you out there Colin?
Igor Kovalenko <> Igor.Kovalenko@motorola.com> > wrote:
Has anyone tried to run TCL 8.x tests on RTP?
It compiles relatively clean, except that configure does not detect that
RTP has fd_set (probably because it’s typedef, rather than #define) and
then fd_mask is undefined (should be int).
But there are so many problems with tests that I’m afraid something is
severely broken. Aside of numerous failing tests (strings, parsing,
float precision, some others) there are tests which can’t be executed at
all (basic, event, io, http, httpold). They simply never return, either
spinning in READY or stuck in SIGWAITINFO forever. And if I’m running
big TCP stack instead of tiny then some tests hang the system
altogether.
The stack test fails with SIGSEGV.
I’m not at all familiar with TCL to look into tests, unfortunately. Any
volunteers to dig the issue?
At one point in time we had someone in-house who had done the full
TCL/Tk port and had managed (if I remember correctly) to get it
to pass through all of it’s tests cleanly. It would be my hope
that our diffs/mods were sent back for inclusion but I’m only
guessing. Colin would know more … you out there Colin?
And who might be ‘him’? I did not try CVS versions but latest public
alpha behaves same way, so either changes weren’t merged or something
changed in OS since then. Also, I’m not sure about all tests, but some
of them certainly would indicate problems with NTO libc. I believe you
guys should want to look into this. Must be very easy to reproduce for
you Colin.
I did think that his changes were merged back into the main tree.
Thomas Fletcher <> thomasf@qnx.com> > wrote:
Igor Kovalenko <> Igor.Kovalenko@motorola.com> > wrote:
Has anyone tried to run TCL 8.x tests on RTP?
It compiles relatively clean, except that configure does not detect that
RTP has fd_set (probably because it’s typedef, rather than #define) and
then fd_mask is undefined (should be int).
But there are so many problems with tests that I’m afraid something is
severely broken. Aside of numerous failing tests (strings, parsing,
float precision, some others) there are tests which can’t be executed at
all (basic, event, io, http, httpold). They simply never return, either
spinning in READY or stuck in SIGWAITINFO forever. And if I’m running
big TCP stack instead of tiny then some tests hang the system
altogether.
The stack test fails with SIGSEGV.
I’m not at all familiar with TCL to look into tests, unfortunately. Any
volunteers to dig the issue?
At one point in time we had someone in-house who had done the full
TCL/Tk port and had managed (if I remember correctly) to get it
to pass through all of it’s tests cleanly. It would be my hope
that our diffs/mods were sent back for inclusion but I’m only
guessing. Colin would know more … you out there Colin?
From Tcl 7.5 on, the source code for Tcl and Tk should compile
and run on Windows machines from the original distribution.
Steve Furr <URL: mailto:furr@qnx.com> reports getting Tcl ported
to QNX without a lot of trouble. He mentions that QNX users who have
the beta X should have gotten a CD-ROM update with Tcl and Tk on the
CD
That’s the last QNXism in there. I submitter a couple of patches to get tk
running and to handle QNX’s //nid/ syntax for 8.2.something.
-Bruce.
Igor Kovalenko wrote:
And who might be ‘him’? I did not try CVS versions but latest public
alpha behaves same way, so either changes weren’t merged or something
changed in OS since then. Also, I’m not sure about all tests, but some
of them certainly would indicate problems with NTO libc. I believe you
guys should want to look into this. Must be very easy to reproduce for
you Colin.
I did think that his changes were merged back into the main tree.
Thomas Fletcher <> thomasf@qnx.com> > wrote:
Igor Kovalenko <> Igor.Kovalenko@motorola.com> > wrote:
Has anyone tried to run TCL 8.x tests on RTP?
It compiles relatively clean, except that configure does not detect that
RTP has fd_set (probably because it’s typedef, rather than #define) and
then fd_mask is undefined (should be int).
But there are so many problems with tests that I’m afraid something is
severely broken. Aside of numerous failing tests (strings, parsing,
float precision, some others) there are tests which can’t be executed at
all (basic, event, io, http, httpold). They simply never return, either
spinning in READY or stuck in SIGWAITINFO forever. And if I’m running
big TCP stack instead of tiny then some tests hang the system
altogether.
The stack test fails with SIGSEGV.
I’m not at all familiar with TCL to look into tests, unfortunately. Any
volunteers to dig the issue?
At one point in time we had someone in-house who had done the full
TCL/Tk port and had managed (if I remember correctly) to get it
to pass through all of it’s tests cleanly. It would be my hope
that our diffs/mods were sent back for inclusion but I’m only
guessing. Colin would know more … you out there Colin?
That was Wayne Collins. He didn’t work here too long, but while he
was here he found quite a few bugs in TCL. And also in our libc.
To my knowledge, all the libc bugs were fixed, and he claimed that the
authors of TCL had received his patches…
And who might be ‘him’? I did not try CVS versions but latest public
alpha behaves same way, so either changes weren’t merged or something
changed in OS since then. Also, I’m not sure about all tests, but some
of them certainly would indicate problems with NTO libc. I believe you
guys should want to look into this. Must be very easy to reproduce for
you Colin.
I did think that his changes were merged back into the main tree.
Thomas Fletcher <> thomasf@qnx.com> > wrote:
Igor Kovalenko <> Igor.Kovalenko@motorola.com> > wrote:
Has anyone tried to run TCL 8.x tests on RTP?
It compiles relatively clean, except that configure does not detect that
RTP has fd_set (probably because it’s typedef, rather than #define) and
then fd_mask is undefined (should be int).
But there are so many problems with tests that I’m afraid something is
severely broken. Aside of numerous failing tests (strings, parsing,
float precision, some others) there are tests which can’t be executed at
all (basic, event, io, http, httpold). They simply never return, either
spinning in READY or stuck in SIGWAITINFO forever. And if I’m running
big TCP stack instead of tiny then some tests hang the system
altogether.
The stack test fails with SIGSEGV.
I’m not at all familiar with TCL to look into tests, unfortunately. Any
volunteers to dig the issue?
At one point in time we had someone in-house who had done the full
TCL/Tk port and had managed (if I remember correctly) to get it
to pass through all of it’s tests cleanly. It would be my hope
that our diffs/mods were sent back for inclusion but I’m only
guessing. Colin would know more … you out there Colin?
Has anyone tried to run TCL 8.x tests on RTP?
It compiles relatively clean, except that configure does not detect that
RTP has fd_set (probably because it’s typedef, rather than #define) and
then fd_mask is undefined (should be int).
But there are so many problems with tests that I’m afraid something is
severely broken. Aside of numerous failing tests (strings, parsing,
float precision, some others) there are tests which can’t be executed at
all (basic, event, io, http, httpold). They simply never return, either
spinning in READY or stuck in SIGWAITINFO forever. And if I’m running
big TCP stack instead of tiny then some tests hang the system
altogether.
The stack test fails with SIGSEGV.
I’m not at all familiar with TCL to look into tests, unfortunately. Any
volunteers to dig the issue?
igor
We have a port of Tcl/Tk 8.3.2 and Blt 2.4u (for Tcl/Tk 8.3) on QRTP.
Normaly this version has a memory leakage due to a missing free.
We have corrected it based on the cvs of the 8.4.
These problems are solved on 8.4 but we didn’t succeed to compile it.
I don’t really know why because I didn’t do it by myself.
We have some little problems with XPhoton, we are talking about these with
Garry Turcote.
Has anyone tried to run TCL 8.x tests on RTP?
It compiles relatively clean, except that configure does not detect that
RTP has fd_set (probably because it’s typedef, rather than #define) and
then fd_mask is undefined (should be int).
But there are so many problems with tests that I’m afraid something is
severely broken. Aside of numerous failing tests (strings, parsing,
float precision, some others) there are tests which can’t be executed at
all (basic, event, io, http, httpold). They simply never return, either
spinning in READY or stuck in SIGWAITINFO forever.
IMHO … this is cause by the ‘ttsetsid() &
ddd/gdb/pdb’ problem, as desribed in the release
notes:
“Signals are not propagated through to the spawned
process even though it has had process-group
control assigned to it”
And if I’m running
big TCP stack instead of tiny then some tests hang the system
altogether.
The stack test fails with SIGSEGV.
This is produced by a recursive script, which is
nested to the maximum nesting depth of 1000. The
test works after setting the maxNestingDepth to
100 and a stack size to 500k … but there must
be something wrong in the memory management.
and a stack size to 500k … but there must