Fix overflow bug in analyzepsbt fee: CAmount instead of int

[sipa]

This causes the fee and estimated_feerate values in the the analyzepsbt output to be off if the amount being spent exceed 21.47483647 BTC.

[promag]

ACK c9963ae, deserves a test?

[DrahtBot]

The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.

Conflicts

Reviewers, this pull request conflicts with the following ones:

• #15508 (Refactor analyzepsbt for use outside RPC code by gwillen)

If you consider this pull request important, please also help to review the conflicting pull requests. Ideally, start with the one that should be merged first.

[achow101]

utACK c9963ae

[jonasschnelli]

utACK c9963ae

[practicalswift]

@sipa Very nice find. Thanks for fixing.

How did you find this issue? Did you find this by manual or tool assisted analysis?

I had this issue in my back log of conversion warnings to investigate manually but you beat me to it :-)

rpc/rawtransaction.cpp:1986: conversion to ‘int’ from ‘CAmount {aka long int}’ may alter its value

FWIW, these are all the related CAmount (int64_t) conversion warnings I have in my back log (I haven't analysed any of these yet):

policy/fees.cpp:1002: conversion to ‘double’ from ‘CAmount {aka long int}’ may alter its value
policy/fees.cpp:1009: conversion to ‘std::set<double>::key_type {aka double}’ from ‘CAmount {aka long int}’ may alter its value
rpc/rawtransaction.cpp:1986: conversion to ‘int’ from ‘CAmount {aka long int}’ may alter its value
test/mempool_tests.cpp:554: conversion to ‘double’ from ‘CAmount {aka long int}’ may alter its value
test/mempool_tests.cpp:558: conversion to ‘double’ from ‘CAmount {aka long int}’ may alter its value
test/mempool_tests.cpp:562: conversion to ‘double’ from ‘CAmount {aka long int}’ may alter its value
test/script_tests.cpp:166: conversion to ‘int’ from ‘CAmount {aka long int}’ may alter its value
test/script_tests.cpp:332: conversion to ‘int’ from ‘CAmount {aka long int}’ may alter its value
txmempool.cpp:1012: conversion to ‘double’ from ‘CAmount {aka long int}’ may alter its value
txmempool.cpp:1013: conversion to ‘double’ from ‘CAmount {aka long int}’ may alter its value
txmempool.h:235: conversion to ‘double’ from ‘CAmount {aka long int}’ may alter its value
txmempool.h:309: conversion to ‘double’ from ‘CAmount {aka long int}’ may alter its value
wallet/test/coinselector_tests.cpp:481: conversion to ‘double’ from ‘CAmount {aka long int}’ may alter its value

utACK c9963ae

[laanwj]

utACK c9963ae

[fanquake]

Backported in 232ef63.

[practicalswift]

@sipa Are you willing to share how this bug was found? I think it would be valuable to do an informal short "post mortem" of this kind of issues.

That would allow us to reason about which mistakes that are recurring and could be candidates for identification by intelligent use of tooling.