installargs loses part of backward
New in the 2017c release, there's a bug in how INSTALLARGS in the Makefile passes the BACKWARD variable down to an inferior make process. Because the definition of INSTALLARGS doesn't put quotes around the value part of BACKWARD, and because BACKWARD is actually (by default) a space-separated list of two files, the inferior make only gets the first file in its BACKWARD value, treating the second file as something to ensure is up to date (which of course it already is). So the inferior make omits one of the BACKWARD files (pacificnew) from whatever it does. The practical effect of this is that when building the binary zone files by, for example, "make posix_only", without having previously done "make tzdata.zi", the US/Pacific-New link is omitted. It's missing from the directory of binary outputs, and missing from the tzdata.zi file that is created as a side effect of building them (thus affecting anything that later tries to use tzdata.zi). That's the only zone name affected, because it's the only zone name defined in the lost source file. If "make tzdata.zi" is done first then that will create a correct tzdata.zi, and subsequent building of the binary zone files will then include the link. This was spotted by my automation, which sanity checks the directory of generated binary zone files against a list of zone names that it extracts from the source files. I think this a sensible check, and that it ought to pass. The fix is that the definition of INSTALLARGS should quote $(BACKWARD), the same way it quotes $(PACKRATDATA). Or perhaps it should not set BACKWARD at all, just like it doesn't set YDATA et al. (But perhaps it should *also* set YDATA et al.) -zefram
Thanks for reporting the problem. The attached patch should fix the issue that you mentioned. I installed it into the development version. Please give it a try. The YDATA issue is separate, and is not addressed by this patch. To some extent I'd rather let that sleeping dog lie, if it's not bothering anybody.
Paul Eggert wrote:
Thanks for reporting the problem. The attached patch should fix the issue that you mentioned. I installed it into the development version. Please give it a try.
Works for me. Do you intend to make a quick 2017d for this issue? I'd like to know whether to fudge 2017c manually or to wait. -zefram
On 2017-10-23 13:07, Zefram wrote:
Paul Eggert wrote:
Thanks for reporting the problem. The attached patch should fix the issue that you mentioned. I installed it into the development version. Please give it a try.
Works for me.
Do you intend to make a quick 2017d for this issue? I'd like to know whether to fudge 2017c manually or to wait.
Forward note from OP on tz-announce and tz and wait until we need a 2017d? Shouldn't be long as the politicans are back from their vacations ;^> Make a new release only if major complaints of missing US/Pacific-New are heard? -- Take care. Thanks, Brian Inglis, Calgary, Alberta, Canada
participants (3)
-
Brian Inglis -
Paul Eggert -
Zefram