The json test data serialized in hex can be quite large and would be parsed by the compiler each time the header file is included, causing OOM even with make -j 1.
Slightly decrease the memory usage by compiling the json to hex separately from the tests and do it at most once (regardless of how often the header is included).
I couldn't find a difference in memory usage, so this should be closed.
For reference:
| file | rss (kB) | compile time (s) |
|---|---|---|
| test/data/base58 encode decode.json hex.cpp | 48588 | 0.98 |
| test/data/blockfilters.json hex.cpp | 52312 | 0.73 |
| test/data/key io invalid.json hex.cpp | 49504 | 0.84 |
| test/data/key io valid.json hex.cpp | 51484 | 0.70 |
| test/data/script tests.json hex.cpp | 106008 | 2.04 |
| test/data/sighash.json hex.cpp | 105196 | 1.88 |
| test/data/tx invalid.json hex.cpp | 61392 | 0.90 |
| test/data/tx valid.json hex.cpp | 67280 | 1.19 |
@MarcoFalke What tools do you use to construct the nice rss/compile time tables? Do you measure using time?