Tcl Source Code

OriginalBugID: 6347 Bug
Version: 8.4a1
SubmitDate: '2000-10-09'
LastModified: '2000-10-25'
Severity: MED
Status: UnAssn
Submitter: techsupp
OS: Windows 98
FixedDate: '2000-10-25'
ClosedDate: '2000-10-25'


Name:
David Cuthbert

Comments:
Opening rmd.bat and mkd.bat in a text editor and resaving them usually
    works.

ObservedBehavior:
Cannot build from the tcl8.4a1 ZIP file (using Win98, MSVC 5 and makefile.vc).  The Makefile fails to create the Release and Debug directories because the mkd.bat and rmd.bat scripts were saved with Unix linefeeds.  (Symptom: Make echos the word "OFF" twice, tries to compile the first file, and fails)

Logged In: YES 
user_id=72656

fixed in 8.4a4cvs.

Logged In: YES 
user_id=7549

Set http://sourceforge.net/tracker/?func=detail&aid=451333&group_id=10894&atid=360894

to fixed/closed when this is fixed/closed (patch accepted and comitted), too.

Logged In: YES 
user_id=7549

Should this move over to the patches tracker?

File Added - 11908: patch.txt

Logged In: YES 
user_id=7549

Mo, this is a build/makedist issue, not a repository 
issue. -kb will twist the problem into knots.

Patch attached for unix/Makefile.in .  I know the script 
works, but am unable to test it, as I don't run unix here.

Logged In: YES 
user_id=90858

I don't think this problem is fixed. I fixed it by adding
the -kb flag to the files but davygrvy undid my fix and
left it in the old state. This bug is now assigned to him
so he should deal with it.

Logged In: YES 
user_id=72656

This was marked 'Fixed' but 'Open', but I believe it was 
meant to be closed.  In any case, perhaps the right 
solution is to run tools/fixEOL.tcl on the necessary files 
before we make the distro.

Logged In: YES 
user_id=79902

Would the attached script help resolve this more sensibly?

File Added - 10579: fixnl.tcl

Logged In: YES 
user_id=7549

Use of the -kb on a text file is improper, your just going 
to twist yourself in a knot.

Say I do a checkout on Mac and edit the file and my editor 
decided to strip off the LF.  Then I commited.

haha.

you'll be revisiting this one day!

Logged In: YES 
user_id=90858

I changed the files to binary with CRLF linefeeds.
That should fix things, if not this bug report
can be re-opened.

Logged In: YES 
user_id=72656

Go ahead with the -k flag.

Logged In: YES 
user_id=7549

I find that treating CVS's behavior of <EOL> is perfect for 
{editing mode}, but building and running should require 
some added work in the makefiles to ensure proper <EOL> on 
the special files.  See 219217.

One cannot guarentee what <EOL> will be upon checkout.  Fix 
it in the makefiles.

Logged In: YES 
user_id=90858

Well, perhaps we could check the files into CVS
using the -k flag. That would treat the file
as binary so that \r\n would be used as EOL
even when checked out on a UNIX system.

Logged In: YES 
user_id=7549

I beleive the source releases are made from the unix 
makefile, and thus the cvs checkout was done with a unix 
cvs client which sets \n as <eol>.  IMO, this is a TEA 
misunderstanding which states all source files should in LF 
mode.  Some just can't be in LF.  Same is true for the .ini 
file used for the help compiler (CRLF or death).