diff|patch

diff|patch

[Vadim]

From: Scott Schwartz <schwartz@galapagos.cse.psu.edu>

>I don't get it.

You're not alone.

>As far as I can tell, context diffs and patch are not
>a matter of unix compatability, creeping features, religious wars, or
>anything like that.  It's just a fact that without patch applying these
>updates is intolerably arduous, and without context diffs you have to
>manually examine the files, split them into sections for patch, and
>copy new files into place, and suffer a greater likelihood of error.
>Why do it the hard way, when simple, effective, minimalistic, software
>tools are available to do the job?

diff with context is undoubtly convinient.  As for the patches,
may i propose that all patches be published in format like that:

	ed - /sys/src/cmd/mycmd.c << '/&'
	.... output of diff -e
	/&

and applied only to the CD-ROM code (i.e. no patches on patches).

(For you computer history buffs -- /& was the terminating card in datasets
in DOS/360.  OS/370 and later used '//'.  //GO.SYSIN DD * was used
as indicator of a beginning of the dataset following this card, for
step GO (program run). '*' means that the data is inlined, the end
of data was indicated by any card containting // in first two columns.
So, the bundle's usage of the sacred JCL is revisionist :)

--vadim

diff|patch

[Sandy]

On Mon, 21 Aug 1995 05:08:24 -0400, 
Vadim Antonov  <avg@postman.ncube.com> wrote:

>As for the patches,
>may i propose that all patches be published in format like that:
>
>	ed - /sys/src/cmd/mycmd.c << '/&'
>	.... output of diff -e
>	/&

Sounds sensible.

>and applied only to the CD-ROM code (i.e. no patches on patches).

Yes! Absolutely.
 --
 Sandy Harris
 sharris@fox.nstn.ca

diff|patch

[forsyth]

>>anything like that.  It's just a fact that without patch applying these
>>updates is intolerably arduous, and without context diffs you have to
>>manually examine the files, split them into sections for patch, and

based on my experience over the years with diff|patch,
i am reluctant to apply patch to anything i care about and not
manually examine the files before and after, especially with patches coming
from a number of sources applied to (possibly) different base versions.

i'd be happier receiving a package that contained (say) the checksums of
the original files on the originating site, a set of diff -e changes,
and a script that checked before applying
that my copy of the source was the right version and also checks
for instance that the output file matches the new version
on the originating site (it might suffice to check
that diff applied to input and output files reproduces the diff output
in the package).  if precondition or postcondition fails, then i really do want to know about it.

diff|patch

[Scott]

I don't get it.  As far as I can tell, context diffs and patch are not
a matter of unix compatability, creeping features, religious wars, or
anything like that.  It's just a fact that without patch applying these
updates is intolerably arduous, and without context diffs you have to
manually examine the files, split them into sections for patch, and
copy new files into place, and suffer a greater likelihood of error.
Why do it the hard way, when simple, effective, minimalistic, software
tools are available to do the job?