ANNOUNCE: "Compiler 1.3" is available
Andreas Otto
Dear Customer,
"Compiler" 1.3, a tool for creating native binaries ot of Tcl code
is available for download.
Please refer to:
http://www.compiler-factory.com/01_04_16-press_E.html
for the public press release
Your Sincerely
aotto
--
================================================================
(C) Compiler-Factory Phone: ++49-(0)8152-399540
Dipl.-Ing Andreas Otto mailto:in...@compiler-factory.com
Business Solutions http://www.compiler-factory.com
Ulmenstrasse 3 => "Compiler", FastWeb, OpMenu
D-34289 Zierenberg => C, C++, Tcl, HTML, database,
=================================================================
Jason Kane
(my system: x386 Linux, Redhat 7.0)
I just puts myself through 10 rounds against this one. My first problem
was that it assumes /usr/bin/ksh is a valid shell. Perhaps I'm the
oddball here, but wouldn't it be a bit more friendly to use bash? This
is particularly strange since the install instructions tell you to
modify your ~/.bashrc anyway.
So, downloaded, compiled and installed pdksh (public domain variant of
ksh).. no problem. Now I've got to play the environmental variable
game. The install sets "COMPILER_MASTER" to the toplevel dir of the
installation (no prob). The actual program(s) expect "COMPILER_HOME".
Again, not a problem, but it's clear that the installation process given
in the documentation doesn't work. Draw your own conclusions.
Now I figure I'll crank out a test, see what happens. So (hello.tcl):
puts "Hello World!"
exit 0
I figure I'll be nice and explicit with the exit call. So I run
"bin/Tcl2Meta -C hello.tcl". The result?
You run out of time,
please contact <ao...@t-online.de> for a new edition
Abort
Evidently there's a licence key that's either non-existant or out of
date. I recalled seeing a licence file on the site, so I downloaded it
and cracked it open. It looks simple enough, one shell script to run.
Of course, this one doesn't use /usr/bin/ksh, this one uses /bin/ksh.
Crank out a symlink.. and finally run the licence installer. It appears
to work, I hop back and try to run Tcl2Meta again... but it still gives
"You run out of time". TKO, and I'm out. For software with a 1280$
price tag for commercial use (290$ Tcl2Meta + 990$ Meta2C, not much
point to getting the first without the second) I'm pretty leary.
I'm hoping that these installation issues will be resolved, at which
time I'll be sure to post a description of the 'new' installation as
well as some benchmarks of results.
tony summerfelt
>I just puts myself through 10 rounds against this one.
i've been waiting for someone here to actually tackle this program...
i think you just saved people a LOT of valuable time...
> For software with a 1280$
>price tag for commercial use (290$ Tcl2Meta + 990$ Meta2C, not much
>point to getting the first without the second) I'm pretty leary.
don't forget the $60,000 for him to add a new language :/
i'm not above paying for software, but for tcl compilers we have wrap,
freewrap, and tclpro, for free, and they seem to do the job just fine...
--
telnet://ventedspleen.dyndns.org
http://members.home.net/tsummerfelt1
Andreas Otto
> Jason Kane wrote on 27 Apr 2001
>
> >I just puts myself through 10 rounds against this one.
>
> i've been waiting for someone here to actually tackle this program...
>
> i think you just saved people a LOT of valuable time...
>
> > For software with a 1280$
> >price tag for commercial use (290$ Tcl2Meta + 990$ Meta2C, not much
> >point to getting the first without the second) I'm pretty leary.
>
> don't forget the $60,000 for him to add a new language :/
>
> i'm not above paying for software, but for tcl compilers we have wrap,
> freewrap, and tclpro, for free, and they seem to do the job just fine...
>
I allways recommend the people to use free software if possible
( I just do it too )
but if they reach the borders of free software they should have a
way to go
mfg
aotto :)
Andreas Otto
> Andreas Otto wrote:
>
> > Dear Customer,
> >
> > "Compiler" 1.3, a tool for creating native binaries ot of Tcl code
> > is available for download.
> >
> > Please refer to:
> >
> > http://www.compiler-factory.com/01_04_16-press_E.html
> >
> > for the public press release
> >
> > Your Sincerely
> >
> > aotto
> > --
> > ================================================================
> > (C) Compiler-Factory Phone: ++49-(0)8152-399540
> > Dipl.-Ing Andreas Otto mailto:in...@compiler-factory.com
> > Business Solutions http://www.compiler-factory.com
> > Ulmenstrasse 3 => "Compiler", FastWeb, OpMenu
> > D-34289 Zierenberg => C, C++, Tcl, HTML, database,
> > =================================================================
>
> (my system: x386 Linux, Redhat 7.0)
>
> I just puts myself through 10 rounds against this one. My first problem
> was that it assumes /usr/bin/ksh is a valid shell. Perhaps I'm the
You are right this is very bad
-> "Compiler" 2.0 has no external dependentcies
> oddball here, but wouldn't it be a bit more friendly to use bash? This
> is particularly strange since the install instructions tell you to
> modify your ~/.bashrc anyway.
>
> So, downloaded, compiled and installed pdksh (public domain variant of
> ksh).. no problem. Now I've got to play the environmental variable
> game. The install sets "COMPILER_MASTER" to the toplevel dir of the
> installation (no prob). The actual program(s) expect "COMPILER_HOME".
The current installation is split:
"COMPILER_MASTER" is the master ( you never work with the master )
this part you already download
"COMPILER_HOME" is your current project
this part you have to download
to see how to setup your own project have a look to the examples
or use the empty project template.
All this will disappear with "Compiler" 2.0
> Again, not a problem, but it's clear that the installation process given
> in the documentation doesn't work. Draw your own conclusions.
>
> Now I figure I'll crank out a test, see what happens. So (hello.tcl):
>
> puts "Hello World!"
> exit 0
>
> I figure I'll be nice and explicit with the exit call. So I run
> "bin/Tcl2Meta -C hello.tcl". The result?
never use the low-level interface directly,
never work in the MASTER
this are the core bug the most people do.
There is a step by step example in
README.compile
which help you based on the TclEX example
The low-level interface is for people who realy know what they do
and should not be used for the first contact with "Compiler".
In "Compiler" 2.0 the Low-level Interface will disappear because
it produce more confusion than help.
>
> You run out of time,
> please contact <ao...@t-online.de> for a new edition
> Abort
the message appears if you use compiler without installing the
client: "Compiler" expect a valid licence key in OT_KEY_LIB which points to
the dynamic library at :
OT_KEY_LIB=$COMPILER_HOME/lib/key/libOtKey1.3.so
export OT_KEY_LIB
>
> Evidently there's a licence key that's either non-existant or out of
> date. I recalled seeing a licence file on the site, so I downloaded it
> and cracked it open. It looks simple enough, one shell script to run.
> Of course, this one doesn't use /usr/bin/ksh, this one uses /bin/ksh.
> Crank out a symlink.. and finally run the licence installer. It appears
> to work, I hop back and try to run Tcl2Meta again... but it still gives
> "You run out of time". TKO, and I'm out. For software with a 1280$
> price tag for commercial use (290$ Tcl2Meta + 990$ Meta2C, not much
> point to getting the first without the second) I'm pretty leary.
>
> I'm hoping that these installation issues will be resolved, at which
> time I'll be sure to post a description of the 'new' installation as
> well as some benchmarks of results.
>
Thank you for trying it
i hope "Compiler" 2.0 will fit your needs
mfg
aotto :)
Jason Kane
usage is as simple as prowrap (not exactly a low-bar, prowrapping a complex
package is non-trivial) the compiler product (assuming it operates as
advertised) fills a sizable niche; particularly in non-commercial software
(assuming the resulting executables consume less disk space than the
prowrapped version of the same tcl script, which should be an easy
assumption).
Andreas Otto wrote:
> The current installation is split:
>
> "COMPILER_MASTER" is the master ( you never work with the master )
> this part you already download
> "COMPILER_HOME" is your current project
> this part you have to download
...
>
>
> There is a step by step example in
>
> README.compile
>
> which help you based on the TclEX example
> The low-level interface is for people who realy know what they do
> and should not be used for the first contact with "Compiler".
...
Jason Kane
command line scripts (no Tk) seem to work, but there's some excessive
screen barf. Needed to have a valid init.tcl hanging around.
Tk script didn't work, but I didn't really try very hard.
the executable version of a lame benchmark is significantly faster than
the 'pure' tcl version.
)
Second Attempt at producing an executable version of the tcl script:
puts "Hello World!"
exit 0
using Andreas Otto's "Compiler 1.3" product.
(note: this is the entire process I went through. When properly done, it
should certainly be much faster and simpler. There is a Project template on
the compiler web site that would probably have been a good idea to use. I
didn't really notice it until I was browing around the site during step #27.)
Steps (including all my screwups):
1) created a new directory to hold my 'project'
2) created 'build' subdirectory
3) created hello.tcl in project directory
4) created TMake.files in build subdirectory (echo "hello" > TMake.files)
5) tried TMake, it assumes /bin/tclsh (no big deal), symlink /usr/bin/tclsh to
/bin/tclsh
6) tried Tmake, it returns "couldn't read file "/tools/tmk/src/tmk": no such
file or directory"
7) accoring to tmk's README the tmk's src directory needs to be in the system
search path. I figure I'll just do it right. Moved Compiler.CVS tree into
/usr/local/sbin/Compiler.CVS, added /usr/local/sbin/Compiler.CVS/tools/tmk/src
to my path, added the /bin while I'm at it.
8) added: eval `/usr/local/sbin/Compiler.CVS/.Init.sh` to my .bashrc
9) same error as #6, the tmk docs suggest adding a TMK_HOME variable. Why
not..
10) didn't help. Ran an strace TMake to see where it's looking.. well, duhh.
it wants to be hanging off the root. Ours is not to reason why... hanging
tools off my root..
11) I mv'd tools, and was rewarded with :
warning: main configuration file
/usr/local/sbin/Compiler.CVS/tools/tmk/config/config.tmk missing
12) changing the mv to a symlink..
13) getting closer, got some screen-barf, the relevant part is:
tmk: in directory /home/jkane/compiler/Project.CVS
tmk: could not read TMakefile
tmk: exiting.
14) so I'm copying Project.CVS/build/TMake.file to Projects.CVS/TMakefile
15) that was wrong, I got
tmk: in directory /home/jkane/compiler/Project.CVS
invalid command name "hello"
while executing
"hello
16) which is at least a tcl error. But I'm missing something important. The
README.compile refers to a whole directory structure in my Project directory
that doesn't exist. I'm guessing there is a template somewhere for the proper
structure? Looking around... there's a template subdir under Compiler, but it
doesn't look very helpful. There's a build directory, so I'm copying that
into my project directory.
17) The example file TclEx refered to in the README doesn't exist, but
TMake.files.tpl does. Lo and behold, it looks like I made a mistake up on #4,
it should have been : echo "OtFile hello" > TMake.files (found by looking at
TMake.files.tpl). Try this again..
18) yeah! looks like we're getting somewhere. After mv'ing the /src
directory underneith the build directory (and putting my hello.tcl into the
src directory), I run "TMake tcl.all" and get:
tmk: in directory /home/jkane/compiler/Project.CVS/build
removing .......... m_hello
removing .......... t_hello
removing .......... t_exe
/usr/local/sbin/Compiler.CVS/bin/Tcl2Meta[31]: Error: not found
/usr/local/sbin/Compiler.CVS/bin/Tcl2Meta[43]: COMPILER_HOME: parameter not
set
Tcl2Meta() - ERROR: you have to set the input file <-C>.
Please try "Tcl2Meta -h" for additional information.
19) tried: export COMPILER_HOME=/usr/local/sbin/Compiler.CVS/
results (TMake tcl.is):
You run out of time,
please contact <ao...@t-online.de> for a new edition
Aborted
20) but this was mentioned in the newsgroup post, so lets see here... ah, a
simple: (added to .bashrc)
export OT_KEY_LIB="/usr/local/sbin/Compiler.CVS/lib/key/libOtKey1.3.so
21) mkdir var; mkdir var/meta; mkdir var/ETcl ... there's a problem here.
copying the Compiler.CVS/var to projects (and lib), moved everything in build
down to the Project directory. Running TMake tcl.all gives:
Tcl2Meta() - ERROR: File</home/jkane/compiler/Project.CVS/exe/i686/Tcl2Meta>
not available
22) copying the exe tree into the project.. better:
- could not build i_setup from /home/jkane/compiler/Project.CVS/.ENV.local:
no rule how to make non-existing /home/jkane/compiler/Project.CVS/.ENV.local
23) there's a .ENV.local.tmpl, I'll tweak it and rename it to .ENV.local; now
I get
could not build t_exe from
/home/jkane/compiler/Project.CVS/template/File.tpl: no rule how to make
non-existing /home/jkane/compiler/Project.CVS/template/File.tpl
24) so I dupe the template directory.. mucho closer:
tmk: couldn't open "/home/jkane/compiler/Project.CVS/var/Exe/HelloWorld": no
such file or directory
tmk: exiting.
25) the var/Exe directory doesn't exist, lets see what happens when we create
it.
26) No errors! Looks like it worked. Trying "TMake c.all" ... :
sh: /home/jkane/compiler/Project.CVS/build/TMake.bin/ot_create_proto.ksh: No
such file or directory
27) looks like there is a reason for the build subdir after all. Moving
TMake.bin to build/TMake.bin doesn't work. Lets try a symlink.. hey, looks
good. I saw GCC compiling and it looks like we've got an executable. Now to
find it..
it's /var/Env/HelloWorld, weighing it at a hefty 640,335 (570,084 stripped).
Not bad compared to a prowrap. But does it work? Running HelloWorld gives:
application-specific initialization failed: Can't find a usable init.tcl in
the following directories:
/home/jkane/compiler/Project.CVS/var/Env/lib
/home/dev1usr/Project/XPress2.CVS/template/C-Tmpl/lib/tcl/lib/tcl8.3
/home/jkane/compiler/Project.CVS/var/lib/tcl8.3
/home/jkane/compiler/Project.CVS/lib/tcl8.3
/home/jkane/compiler/Project.CVS/lib/tcl8.3/library
/home/jkane/compiler/Project.CVS/library
/home/jkane/compiler/Project.CVS/tcl8.3/library
/home/jkane/compiler/tcl8.3/library
/home/dev1usr/Project/XPress2.CVS/template/C-Tmpl/lib/tcl/lib/tcl8.3
I think I read something about this in the FAQ on the compiler homepage.. not
the FAQ I guess it was somewhere else.
grep -r "init.tcl" * to get a little more information. That was useless,
maybe in the compiler directory.. doesn't help. We don't want to die here..
so we'll stuff a legit init.tcl into the first directory checked
(Project.CVS/var/Env/lib). Ok, output of HelloWorld (executable) is now:
[jkane@kane Env]$ ./HelloWorld
HelloWorld(v0.2) ............ BUILD(3)
RunTime(v1.3) ............... BUILD(640)
OtKey(v1.3) ................. BUILD(1)
This is the "default" licence Key of "Compiler" :
"Compiler" is *not* free and belongs to
Dipl.-Ing. Andreas Otto
Ulmenstrasse 3
D-34289 Zierenberg
Germany
ao...@t-online.de
Using is permitted to free software only.
Commercial projects need a business license.
Have a look to:
http://www.compiler-factory.com
for additional information.
/* the key is valid until end of MAY 2001 */
Hello world!
------------------
Finally! That's a lot of screen barf though.. presumably the 'commercial'
version allows one to get rid of it. Now that I have a 'working' .TCL -->
.EXE (pardon the somewhat absurd use of the dossy extention convention) So
lets mess with it.
What happens when we replace the 'old' hello.tcl with:
label .hello -text "Hello world!"
pack .hello
TMake tcl.all didn't have any problems, TMake c.all didn't have any problems
either. But HelloWorld now gives:
(same screen barf as before)
application-specific initialization failed: invalid command name "label"
So we change .ENV.local from tclsh to wish and from tcl to tk, then TMake
tcl.all; TMake c.all ... I saw some X11 libraries in there, maybe this
worked.. lets see: Running HelloWorld now pops up an empty widow and returns:
Application initialization failed: invalid command name "tcl_findLibrary".
Bummer.
The filesize on HelloWorld is up to 1,379,501 (1,249,596 stripped) thanks to
the Tk/X11 libraries.
Changing Hello.tcl to:
puts [time {
set cat "Hello world!"
set dog $cat
set cow [list $cat $dog]
set goat(beaver) [list $cat $dog $cow]
} 1000]
exit 0
From the command line tclsh (8.3) I get 20 microseconds per iteration.
from wish (8.3) I also get 20 microseconds per iteration (duh)
from the compiled HelloWorld (Tk) I get: Application initialization failed:
invalid command name "tcl_findLibrary" (stupid me)
from the compiled HelloWorld (Tcl) I get 14 microseconds per iteration.
That's 142% of the tclsh version. Of course this is a far cry from a 'good'
benchmark but it's still reasonably impressive.
-----------
That's all for today folks, I hope this is useful to some of you. It seems to
me like the upcoming version of compiler (2.0) will be a strong alternative to
prowrap for non-commercial code.
tony summerfelt
> 1) created a new directory to hold my 'project'
.
.
.
> 26) No errors! Looks like it worked. Trying "TMake c.all" ... :
i dunno, call me skeptical but 'freewrap test.tcl' worked great for me...
extracting the tarball could be considered a step i guess :)
the last time i had to pound away at getting compiler to work was at the
request of my employer for ms c v3.0 with no documentation. and even then it
wasn't that hard...
Andreas Otto
Hi,
thank you for trying it
-> you done the really really hard way to get anythink working
this is not necessarry but there is my first live rule ==>
"you understand somthing really if you have done all possible errors"
(aotto 01 May 2001)
> (summary for anyone only interested in the results:
> command line scripts (no Tk) seem to work, but there's some excessive
> screen barf. Needed to have a valid init.tcl hanging around.
> Tk script didn't work, but I didn't really try very hard.
try the TkDiff example
> the executable version of a lame benchmark is significantly faster
> than
> the 'pure' tcl version.
well it's a native binary,
i always said that compiled Tcl is the fastes available
script like progrmming language
> )
>
> Second Attempt at producing an executable version of the tcl script:
>
> puts "Hello World!"
> exit 0
>
> using Andreas Otto's "Compiler 1.3" product.
>
> (note: this is the entire process I went through. When properly done, it
> should certainly be much faster and simpler. There is a Project template
> on
> the compiler web site that would probably have been a good idea to use. I
> didn't really notice it until I was browing around the site during step
that's a job for me to get everything more clear and easy to use
You have done everything by hand this mean you programm your own
"Compiler" client this is good if you want to figure out how the internals
are working but really not neseccary.
I always recomment to download an example to make the first step
into the "Compiler" technology
> #27.)
>
> Steps (including all my screwups):
>
> 1) created a new directory to hold my 'project'
>
> 2) created 'build' subdirectory
>
> 3) created hello.tcl in project directory
>
> 4) created TMake.files in build subdirectory (echo "hello" > TMake.files)
>
> 5) tried TMake, it assumes /bin/tclsh (no big deal), symlink
> /usr/bin/tclsh to /bin/tclsh
>
> 6) tried Tmake, it returns "couldn't read file "/tools/tmk/src/tmk": no
> such file or directory"
>
> 7) accoring to tmk's README the tmk's src directory needs to be in the
> system
> search path. I figure I'll just do it right. Moved Compiler.CVS tree
> into /usr/local/sbin/Compiler.CVS, added
> /usr/local/sbin/Compiler.CVS/tools/tmk/src to my path, added the /bin
> while I'm at it.
cat bin/TMake
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
#!/bin/ksh
exec ${COMPILER_HOME}/bin/tclsh ${COMPILER_HOME}/tools/tmk/src/tmk ${1+"$@"}
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
it seems you don't set the COMPILER_HOME environment variable
-> the template does it
>
> 8) added: eval `/usr/local/sbin/Compiler.CVS/.Init.sh` to my .bashrc
right you now set COMPILER_MASTER
Well that's the default licence text
>
> ------------------
>
> Finally! That's a lot of screen barf though.. presumably the 'commercial'
> version allows one to get rid of it. Now that I have a 'working' .TCL -->
> .EXE (pardon the somewhat absurd use of the dossy extention convention)
> So lets mess with it.
>
> What happens when we replace the 'old' hello.tcl with:
>
> label .hello -text "Hello world!"
> pack .hello
>
> TMake tcl.all didn't have any problems, TMake c.all didn't have any
> problems
> either. But HelloWorld now gives:
> (same screen barf as before)
> application-specific initialization failed: invalid command name "label"
>
> So we change .ENV.local from tclsh to wish and from tcl to tk, then TMake
> tcl.all; TMake c.all ... I saw some X11 libraries in there, maybe this
> worked.. lets see: Running HelloWorld now pops up an empty widow and
> returns: Application initialization failed: invalid command name
> "tcl_findLibrary". Bummer.
> The filesize on HelloWorld is up to 1,379,501 (1,249,596 stripped) thanks
this is static linked this is no problem because dynamic linking
is possible to
> to the Tk/X11 libraries.
>
> Changing Hello.tcl to:
>
> puts [time {
> set cat "Hello world!"
> set dog $cat
> set cow [list $cat $dog]
> set goat(beaver) [list $cat $dog $cow]
> } 1000]
>
> exit 0
>
> From the command line tclsh (8.3) I get 20 microseconds per iteration.
> from wish (8.3) I also get 20 microseconds per iteration (duh)
> from the compiled HelloWorld (Tk) I get: Application initialization
> failed: invalid command name "tcl_findLibrary" (stupid me)
> from the compiled HelloWorld (Tcl) I get 14 microseconds per iteration.
> That's 142% of the tclsh version. Of course this is a far cry from a
> 'good' benchmark but it's still reasonably impressive.
If you put a "proc" arround it it will be more impressive.
>
>
> -----------
> That's all for today folks, I hope this is useful to some of you. It
> seems to me like the upcoming version of compiler (2.0) will be a strong
> alternative to prowrap for non-commercial code.
>
--
lvi...@yahoo.com
According to Jason Kane <jk...@broadlink.com>:
:5) tried TMake, it assumes /bin/tclsh (no big deal), symlink /usr/bin/tclsh to
:/bin/tclsh
Actually, this IS a big deal for people on non-stand-alone machines, like
me. NOTHING is installed by me in /bin ; that's where the OS from the
vendor goes. I have no write permission to /bin, /usr/bin, or for that
matter /usr/local/bin! The best bet is for this to be user definable
at install time...
:6) tried Tmake, it returns "couldn't read file "/tools/tmk/src/tmk": no such
:file or directory"
:
:7) accoring to tmk's README the tmk's src directory needs to be in the system
:search path. I figure I'll just do it right. Moved Compiler.CVS tree into
:/usr/local/sbin/Compiler.CVS, added /usr/local/sbin/Compiler.CVS/tools/tmk/src
:to my path, added the /bin while I'm at it.
Again, ALL paths need to be user-definable at install time; otherwise, the
product won't be useable in a commercial environment.
:8) added: eval `/usr/local/sbin/Compiler.CVS/.Init.sh` to my .bashrc
If bash is a requirement, then that should be reconsidered, so that the user
can either use a compatible shell, or at the very least allow the user to
write their own compatible shell and be able to insert it into the process.
:9) same error as #6, the tmk docs suggest adding a TMK_HOME variable. Why
:not..
:
:10) didn't help. Ran an strace TMake to see where it's looking.. well, duhh.
:it wants to be hanging off the root. Ours is not to reason why... hanging
:tools off my root..
Actually, ours IS to question - dependance on installation in root is "evil".
:That's all for today folks, I hope this is useful to some of you. It seems to
:me like the upcoming version of compiler (2.0) will be a strong alternative to
:prowrap for non-commercial code.
:
Assuming that all of the concerns about installation are addressed, and that
a working step by step example of how to get things working is included.
--
--
"See, he's not just anyone ... he's my son." Mark Schultz
<URL: mailto:lvi...@cas.org> <URL: http://www.purl.org/NET/lvirden/>
Even if explicitly stated to the contrary, nothing in this posting