At the moment, the version of org-web-tools in MELPA is (20180920 311). The version that gets built on my system at the moment, which is from a different commit that I just pushed to the package's repo, is (20180920 54). Here's the relevant part of package-archive-contents from the sandbox Emacs:
(org-web-tools #s(package-desc org-web-tools
(20180920 311)
"Display and capture web content with Org-mode"
((emacs
(25 1))
(org
(9 0))
(dash
(2 12))
(esxml
(0 3 4))
(s
(1 10 0))
(request
(0 3 0)))
tar "melpa" nil
((:commit . "30315bcf6183b49352ce37d66615df3ec67f8319")
(:keywords "hypermedia" "outlines" "org" "web")
(:authors
("Adam Porter" . "[email protected]"))
(:maintainer "Adam Porter" . "[email protected]")
(:url . "http://github.com/alphapapa/org-web-tools"))
nil)
#s(package-desc org-web-tools
(20180920 54)
"Display and capture web content with Org-mode"
((emacs
(25 1))
(org
(9 0))
(dash
(2 12))
(esxml
(0 3 4))
(s
(1 10 0))
(request
(0 3 0)))
tar "sandbox" nil
((:commit . "a2b2cdfa75d50dd00f4106dfdb6b6bec19280b7d")
(:keywords "hypermedia" "outlines" "org" "web")
(:authors
("Adam Porter" . "[email protected]"))
(:maintainer "Adam Porter" . "[email protected]")
(:url . "http://github.com/alphapapa/org-web-tools"))
nil))
The problem is that I need to test the newer commit in the sandbox, but the sandbox Emacs insists on installing the older version from MELPA.
I'm guessing this is caused by the build server being in a different time zone.
I don't know of an easy way to fix or work around this problem, and it doesn't happen very often, but at least this documents it.
Thanks.
I think that may be the cause for #5809 too
I don't know of an easy way to fix or work around this problem, and it doesn't happen very often, but at least this documents it.
It would be better to do something about it though, probably in package-build--parse-time.
It would be better to do something about it though, probably in
package-build--parse-time.
I guess we'd want to make the MELPA build server time zone the default in package-build too, and probably set via a variable. Then package-build--parse-time would append it to the string parsed by date-to-time...
What if UTC were always used? That seems like the most correct solution.
We've just switched to UTC.
Most helpful comment
We've just switched to UTC.