Discussion:
master 7b1026c: * make-dist: Don't fail if building --without-makeinfo.
(too old to reply)
Glenn Morris
2018-05-04 15:48:18 UTC
Permalink
branch: master
commit 7b1026cbf48fa8a3f31497910f696a5ba28476cb
[...]
* make-dist: Don't fail if building --without-makeinfo.
I'm not sure that make-dist should allow the creation of tarfiles
without the info files. Almost certainly not by default, anyway.
Eli Zaretskii
2018-05-04 17:36:09 UTC
Permalink
Date: Fri, 04 May 2018 11:48:18 -0400
branch: master
commit 7b1026cbf48fa8a3f31497910f696a5ba28476cb
[...]
* make-dist: Don't fail if building --without-makeinfo.
I'm not sure that make-dist should allow the creation of tarfiles
without the info files. Almost certainly not by default, anyway.
I tend to agree.

Noam, what was the rationale for this change?
Noam Postavsky
2018-05-05 02:15:17 UTC
Permalink
Post by Eli Zaretskii
Post by Glenn Morris
* make-dist: Don't fail if building --without-makeinfo.
I'm not sure that make-dist should allow the creation of tarfiles
without the info files. Almost certainly not by default, anyway.
I tend to agree.
Noam, what was the rationale for this change?
I have the automated tests setup to on my gitlab fork relying on
make-dist. Up until its recent changes [1: 83a6224d60], it succeeded
even though the info files weren't built.

So would adding a --no-info argument (as in the attached) be okay?

[1: 83a6224d60]: 2018-04-09 13:35:15 -0700
make-dist: check exit statuses more carefully
https://git.savannah.gnu.org/cgit/emacs.git/commit/?id=83a6224d607c645cadbe371c921928166da0aef0
Eli Zaretskii
2018-05-05 06:46:11 UTC
Permalink
Date: Fri, 4 May 2018 22:15:17 -0400
I have the automated tests setup to on my gitlab fork relying on
make-dist. Up until its recent changes [1: 83a6224d60], it succeeded
even though the info files weren't built.
So would adding a --no-info argument (as in the attached) be okay?
I think it would. But maybe it will make even more sense to try
building the Info files, and only make the failure to do it non-fatal
with that switch specified, WDYT? Because tests might at some future
time include the Info files, for example.

Glenn?
Paul Eggert
2018-05-05 06:53:41 UTC
Permalink
Post by Eli Zaretskii
Post by Noam Postavsky
So would adding a --no-info argument (as in the attached) be okay?
I think it would. But maybe it will make even more sense to try
building the Info files, and only make the failure to do it non-fatal
with that switch specified, WDYT?
Unless you specify --no-update, make-dist already tries to build the info files.
If that fails make-dist goes ahead anyway; perhaps it shouldn't.
Noam Postavsky
2018-05-05 14:41:42 UTC
Permalink
Post by Paul Eggert
Post by Eli Zaretskii
Post by Noam Postavsky
So would adding a --no-info argument (as in the attached) be okay?
I think it would. But maybe it will make even more sense to try
building the Info files, and only make the failure to do it non-fatal
with that switch specified, WDYT?
Unless you specify --no-update, make-dist already tries to build the info
files. If that fails make-dist goes ahead anyway; perhaps it shouldn't.
For the automated testing builds, I pass '--tests --no-update
--no-changelog' to make-dist.
So the failure I want to avoid is just the part where the info files
get hard-linked into the dist directory (because the test env comes
without makeinfo).
Paul Eggert
2018-05-05 06:56:49 UTC
Permalink
Post by Noam Postavsky
So would adding a --no-info argument (as in the attached) be okay?
If --no-info is given, please also change make-dist so that it does not invoke
'make --question info' or 'make info'. Other than that it looks good; thanks.
(Good catch in getting rid of that '[ -d info ]'.)
Noam Postavsky
2018-05-05 14:47:41 UTC
Permalink
Post by Paul Eggert
Post by Noam Postavsky
So would adding a --no-info argument (as in the attached) be okay?
If --no-info is given, please also change make-dist so that it does not
invoke 'make --question info' or 'make info'.
I think Eli is suggesting the opposite, so I'll hold off on making any
changes until we decide where we're going.
Post by Paul Eggert
(Good catch in getting rid of that '[ -d info ]'.)
I can't really claim credit for that, since I was the one who added it ;)
(in the commit mentioned at the top of this thread)
Eli Zaretskii
2018-05-05 15:34:43 UTC
Permalink
Date: Sat, 5 May 2018 10:47:41 -0400
Post by Paul Eggert
If --no-info is given, please also change make-dist so that it does not
invoke 'make --question info' or 'make info'.
I think Eli is suggesting the opposite
Yes. I think it makes more sense to try and then go ahead if
production of Info files fails. It's not a strong opinion, though.
Paul Eggert
2018-05-05 17:33:40 UTC
Permalink
Post by Eli Zaretskii
Yes. I think it makes more sense to try and then go ahead if
production of Info files fails. It's not a strong opinion, though.
Although that would be OK ordinarily, a user who runs './make-dist --no-info'
does not want info files in the tarball and my point was that make-dist should
not try to build them in that case.
Noam Postavsky
2018-05-13 15:05:16 UTC
Permalink
Post by Paul Eggert
Post by Eli Zaretskii
Yes. I think it makes more sense to try and then go ahead if
production of Info files fails. It's not a strong opinion, though.
Although that would be OK ordinarily, a user who runs './make-dist
--no-info' does not want info files in the tarball and my point was that
make-dist should not try to build them in that case.
Yes, I think I'd rather not see noise from error messages about it either.
Loading...