Revision 792b8e96aacf10c23de05b1f96da3236b20121a9 authored by Boud Roukema on 25 September 2020, 18:33:39 UTC, committed by Boud Roukema on 25 September 2020, 18:33:39 UTC
The convention in Maneage for bracketting in makefiles is (it seems to me) to use $(round brackets), not ${curly brackets}, at the 'make' level, and $$nobracketsnorunderscores for variables intended to pass through to the bash shell; while our convention in elaphrocentre modifies this to use $${curly_brackets_and_allow_underscores} for variables to be passed through to bash scripts. However, sometimes we haven't followed this convention fully. According to the GNU make manual [1], ${curly brackets} are just as valid as $(round brackets) at the 'make' level, but mixing the two types is discouraged. This commit hypothesises that there is a bug in the GNU make ore the GNU bash implemention in macOS/Darwin/xnu that occurs when a colon ':' is included indirectly (through nesting) in a ${curly_bracket}, by improving the consistency of using round brackets in `reproduce/analysis/make/analyse-plot.mk`. (Some locale variables are also explicitly set to English in `reproduce/analysis/bash/run-plots`.) In principle, this commit should not cause any difference in behaviour of the pipeline. However, it might fix a bug in macOS/Darwin/xnu. [1]: https://www.gnu.org/software/make/manual/html_node/Syntax-of-Functions.html#Syntax-of-Functions
1 parent 32578e1
File | Mode | Size |
---|---|---|
reproduce | ||
tex | ||
.dir-locals.el | -rw-r--r-- | 1.3 KB |
.file-metadata | -rw-r--r-- | 9.0 KB |
.gitignore | -rw-r--r-- | 994 bytes |
COPYING | -rw-r--r-- | 34.3 KB |
FAQ | -rw-r--r-- | 863 bytes |
README-hacking.md | -rw-r--r-- | 90.2 KB |
README.md | -rw-r--r-- | 21.7 KB |
TODO | -rw-r--r-- | 4.6 KB |
paper.tex | -rw-r--r-- | 90.4 KB |
project | -rwxr-xr-x | 23.3 KB |
Computing file changes ...