setup 2.892 release candidate - please test

Jon Turney
Sun Jul 15 13:23:00 GMT 2018

A new setup release candidate is available at: (64 bit version)    (32 bit version)

Please test and report any problems here.

Changes compared to 2.891:

- Fix building with libgpg-error 1.28

- Raise a fatal error on a corrupt archive in "install from local 
directory" mode

- Packages listed in the "confirm" dialog are now sorted alphabetically

- Fix creating a bogus current version with an empty version number for 
packages which only have a test version

- Fix ignoring the last package version in all setup.ini files but the 
last one read

- Allow command-line uninstallation of specified orphaned packages

- Dependencies of packages specified to be installed or upgraded on the 
command-line are now upgraded, if required (as we now record these 
packages being in a "no changes requested" state, rather than a "keep" 
state that would cause a conflict)

- Packages which have been upgraded automatically are now correctly 
reported as "automatically added" in the "confirm" dialog,

- mirrors.lst is now fetched using https://

- Literal URLs in dialogs now reference

- Fix a bug (dating from 2003) in saving an empty network proxy password

- Correctly handle 'provides:' and 'conflicts:' for installed packages

- Add a '--no-version-check' option, to suppress the warning that a 
newer version of setup is available

- Don't check setup-version: to warn about potential setup.ini parsing 
problems if a setup-minimum-version: is specified.

- Propagate the exit code of the elevated process, when --wait is used
   Addresses: (partially)

- Properly track the sites providing a given version of a packages (To 
correctly handle local installs where required packages are spread 
between cache directories for more than one mirror, and correctly fall 
back to an alternate mirror when multiple mirrors are used and a problem 
occurs connecting to the first mirror tried)

- Don't use CRLF terminators when writing dependency problems to logfile

- Fix a bug which sometimes caused a wildly incorrect value to be used 
for the expected size download progress was measured against

Again, many thanks to Ken Brown for doing all the hard work.

More information about the Cygwin-apps mailing list