source: trunk/third/perl/README.mint @ 14545

Revision 14545, 9.0 KB checked in by ghudson, 25 years ago (diff)
This commit was generated by cvs2svn to compensate for changes in r14544, which included commits to RCS files with non-trunk default branches.
Line 
1##########################################################################
2# *** README.mint
3##########################################################################
4
5If you want to build perl yourself on MiNT (or maybe on an Atari without
6MiNT) you may want to accept some advice from somebody who already did it...
7
8There was a perl port for Atari ST done by ++jrb bammi@cadence.com.
9This port tried very hard to build on non-MiNT-systems.  For the
10sake of efficiency I've left this way.  Yet, I haven't removed bammi's
11patches but left them intact.  Unfortunately some of the files that
12bammi contributed to the perl distribution seem to have vanished?
13
14So, how can you distinguish my patches from bammi's patches?  All of
15bammi's stuff is embedded in "#ifdef atarist" preprocessor macros.
16My MiNT port uses "#ifdef __MINT__" instead (and unconditionally
17undefines "atarist".  If you want to continue on bammi's port, all
18you have to do is to swap the "-D" and "-U" switches for "__MINT__" 
19and "atarist" in the variable ccflags.
20
21However, I think that my version will still run on non-MiNT-systems
22provided that the user has a Eunuchs-like environment (i.e. the
23standard envariables like $PATH, $HOME, ... are set, there is a
24POSIX compliant shell in /bin/sh, and...)
25
26Known problems
27==============
28
29The problems you may encounter when building perl on your machine
30are most probably due to deficiencies in MiNT resp. the Atari
31platform in general.
32
33First of all, if you have less than 8 MB of RAM you shouldn't
34even try to build Perl yourself.  Better grab a binary pre-compiled
35version somewhere.  Even if you have more memory you should take
36some care.  Try to run in a fresh environment (without memory
37fragmented too much) with as few daemons, accessories, xcontrol
38modules etc. as possible.  If you run some AES you should
39consider to start a console based environment instead.
40
41A problem has been reported with sed.  Sed is used to create
42some configuration files based on the answers you have given
43to the Configure script.  Unfortunately the Perl Configure script
44shows sed on MiNT its limits.  I have sed 2.05 with a stacksize
45of 64k and I have encountered no problems.  If sed crashes
46during your configuration process you should first try to
47augment sed's stacksize:
48
49        fixstk 64k /usr/bin/sed
50
51(or similar).  If it still doesn't help you may have a look
52which other versions of sed are installed on your system.
53If you have a KGMD 1.0 installation you will find three
54in /usr/bin.  Have a look there.
55
56Perl has some "mammut" C files.  If gcc reports "internal
57compiler error: program cc1 got fatal signal 10" this is very
58likely due to a stack overflow in program cc1. Find cc1
59and fix its stack.  I have made good experiences with
60
61        fixstk 2 cc1
62
63This doesn't establish a stack of 2 Bytes only as you might
64think.  It really reserves one half of the available memory
65for cc1's stack.  A setting of 1 would reserve the entire
66memory for cc1, 3 would reserve three fourths.  You will have
67to find out the value that suits to your system yourself.
68
69To find out the location of the program `cc1' simply type
70`gcc --print-prog-name cc1' at your shell prompt.
71
72Now run make (maybe "make -k").  If you get a fatal signal 10
73increase cc1's stacksize, if you run out of memory you should
74either decrease the stacksize or follow some more hints:
75
76Perl's building process is very handy on machines with a lot
77of virtual memory but may result in a desaster if you are short
78of memory.  If gcc fails to compile many source files you should
79reduce the optimization.  Grep for "optimize" in the file
80config.sh and change the flags.
81
82If only several huge files cause problems (actually it is not a
83matter of the file size resp. the amount of code but depends on
84the size of the individual funtions) it is useful to bypass
85the make program and compile these files directly from the
86command line.  For example if you got something like the
87following from make:
88
89        CCCMD = gcc -DPERL_CORE ....
90        ...
91        ...: virtual memory exhausted
92
93you should hack into the shell:
94
95        gcc -DPERL_CORE ... toke.c
96
97Please note that you have to add the name of the source file
98(here toke.c) at the end.
99
100If none of this helps, you're helpless.  Wait for a binary
101release.  If you have succeded you may encounter another problem
102at the linking process.  If gcc complains that it can't find
103some libraries within the perl distribution you probably have
104an old linker.  If it complains for example about "file not
105found for xxx.olb" you should cd into the directory in
106question and
107
108        ln -s libxxx.a xxx.olb
109
110This will fix the problem.
111
112This version (5.00402) of perl has passed most of the tests on my system:
113
114Failed Test  Status Wstat Total Fail  Failed  List of failed
115------------------------------------------------------------------------------
116io/pipe.t                    10    2  20.00%  7, 9
117io/tell.t                    13    1   7.69%  12
118lib/complex.t               762   13   1.71%  84-85, 248-251, 257, 272-273,
119                                              371, 380, 419-420
120lib/io_pipe.t                10    1  10.00%  9
121lib/io_tell.t                13    1   7.69%  12
122op/magic.t                   30    2   6.67%  29-30
123Failed 6/152 test scripts, 96.05% okay. 20/4359 subtests failed, 99.54% okay.
124
125Pipes always cause problems with MiNT, it's actually a surprise that
126most of the tests did work.  I've got no idea why the "tell" test failed,
127this shouldn't mean too big a problem however.
128
129Most of the failures of lib/complex seem to be harmless, actually errors
130far right to the decimal point...  Two failures seem to be serious:
131The sign of the results is reversed.  I would say that this is due
132to minor bugs in the portable math lib that I compiled perl with.
133
134I haven't bothered very much to find the reason for the failures
135with op/magic.t and op/stat.t.  Maybe you'll find it out.
136
137##########################################################################
138
139Another possible problem may arise from the implementation of  the "pwd"
140command.  It happened to add a carriage return and newline to its output
141no matter what the setting of $UNIXMODE is.  This is quite annoying since many
142library modules for perl take the output of pwd, chop off the
143trailing newline character and then expect to see a valid path in
144that.  But the carriage return (last but second character!) isn't
145chopped off.  You can either try to patch all library modules (at
146the price of performance for the extra transformation) or you can
147use my version of pwd that doesn't suffer from this deficiency.
148
149The fixed implementation is in the mint subdirectory.  Running
150"Configure" will attempt to build and install it if necessary
151(hints/mint.sh will do this work) but you can build and install it
152explicitly by:
153
154        cd mint
155        make install
156
157This is the fastest solution.
158
159Just in case you want to go the hard way: perl won't even build with a
160broken pwd!  You will have to fix the library modules
161(ext/POSIX/POSIX.pm, lib/Cwd.pm, lib/pwd.pl) at last after building
162miniperl.
163
164A major nuisance of current MiNTLib versions is the implementation
165of system() which is far from being POSIX compliant.  A real system()
166should fork and then exec /bin/sh with its argument as a command
167line to the shell.  The MiNTLib system() however doesn't expect
168that every user has a POSIX shell in /bin/sh.  It tries to work
169around the problem by forking and exec'ing the first token in its argument
170string.  To get a little bit of compliance to POSIX system() it
171tries to handle at least redirection ("<" or ">") on its own
172behalf. 
173
174This isn't a good idea since many programs expect that they can
175pass a command line to system() that exploits all features of a
176POSIX shell.  If you use the MiNTLib version of system() with
177perl the Perl function system() will suffer from the same deficiencies.
178
179You will find a fixed version of system() in the mint subdirectory.
180You can easily insert this version into your system libc:
181
182        cd mint
183        make system.o
184        ar r /usr/lib/libc.a
185        ranlib /usr/lib/libc.a
186
187If you are suspicious you should either back up your libc before
188or extract the original system.o from your libc with
189"ar x /usr/lib/libc.a system.o".  You can then backup the system.o
190module somewhere before you succeed.
191
192Anything missing?  Yep, I've almost forgotten... 
193No file in this  distribution without a fine saying.  Take this one:
194
195        "From a thief you should learn: (1) to work at night;
196        (2) if one cannot gain what one wants in one night to
197        try again the next night; (3) to love one's coworkers
198        just as thieves love each other; (4) to be willing to
199        risk one's life even for a little thing; (5) not to
200        attach too much value to things even though one has
201        risked one's life for them - just as a thief will resell
202        a stolen article for a fraction of its real value;
203        (6) to withstand all kinds of beatings and tortures
204        but to remain what you are; and (7) to believe your
205        work is worthwhile and not be willing to change it."
206
207                        -- Rabbi Dov Baer, Maggid of Mezeritch
208
209OK, this was my motto while working on Perl for MiNT, especially rule (1)...
210
211Have fun with Perl!
212
213Guido Flohr
214--
215mailto:gufl0000@stud.uni-sb.de
216http://stud.uni-sb.de/~gufl0000
Note: See TracBrowser for help on using the repository browser.