nix-super/src/libstore/worker-protocol.hh
John Ericson 37fca662b0 Make KeyedBuildResult, BuildResult like before, and fix bug another way
In https://github.com/NixOS/nix/pull/6311#discussion_r834863823, I
realized since derivation goals' wanted outputs can "grow" due to
overlapping dependencies (See `DerivationGoal::addWantedOutputs`, called
by `Worker::makeDerivationGoalCommon`), the previous bug fix had an
unfortunate side effect of causing more pointless rebuilds.

In paticular, we have this situation:

1. Goal made from `DerivedPath::Built { foo, {a} }`.

2. Goal gives on on substituting, starts building.

3. Goal made from `DerivedPath::Built { foo, {b} }`, in fact is just
   modified original goal.

4. Though the goal had gotten as far as building, so all outputs were
   going to be produced, `addWantedOutputs` no longer knows that and so
   the goal is flagged to be restarted.

This might sound far-fetched with input-addressed drvs, where we usually
basically have all our goals "planned out" before we start doing
anything, but with CA derivation goals and especially RFC 92, where *drv
resolution* means goals are created after some building is completed, it
is more likely to happen.

So the first thing to do was restore the clearing of `wantedOutputs` we
used to do, and then filter the outputs in `buildPathsWithResults` to
only get the ones we care about.

But fix also has its own side effect in that the `DerivedPath` in the
`BuildResult` in `DerivationGoal` cannot be trusted; it is merely the
*first* `DerivedPath` for which this goal was originally created.

To remedy this, I made `BuildResult` be like it was before, and instead
made `KeyedBuildResult` be a subclass wit the path. Only
`buildPathsWithResults` returns `KeyedBuildResult`s, everything else
just becomes like it was before, where the "key" is unambiguous from
context.

I think separating the "primary key" field(s) from the other fields is
good practical in general anyways. (I would like to do the same thing
for `ValidPathInfo`.) Among other things, it allows constructions like
`std::map<Key, ThingWithKey>` where doesn't contain duplicate keys and
just precludes the possibility of those duplicate keys being out of
sync.

We might leverage the above someday to overload `buildPathsWithResults`
to take a *set* of return a *map* per the above.

-----

Unfortunately, we need to avoid C++20 strictness on designated
initializers.

(BTW
https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2021/p2287r1.html
this offers some new syntax for this use-case. Hopefully this will be
adopted and we can eventually use it.)

No having that yet, maybe it would be better to not make
`KeyedBuildResult` a subclass to just avoid this.

Co-authored-by: Robert Hensing <roberth@users.noreply.github.com>
2023-04-15 11:01:31 -04:00

201 lines
5.4 KiB
C++

#pragma once
///@file
#include "store-api.hh"
#include "serialise.hh"
namespace nix {
#define WORKER_MAGIC_1 0x6e697863
#define WORKER_MAGIC_2 0x6478696f
#define PROTOCOL_VERSION (1 << 8 | 35)
#define GET_PROTOCOL_MAJOR(x) ((x) & 0xff00)
#define GET_PROTOCOL_MINOR(x) ((x) & 0x00ff)
/**
* Enumeration of all the request types for the "worker protocol", used
* by unix:// and ssh-ng:// stores.
*/
typedef enum {
wopIsValidPath = 1,
wopHasSubstitutes = 3,
wopQueryPathHash = 4, // obsolete
wopQueryReferences = 5, // obsolete
wopQueryReferrers = 6,
wopAddToStore = 7,
wopAddTextToStore = 8, // obsolete since 1.25, Nix 3.0. Use wopAddToStore
wopBuildPaths = 9,
wopEnsurePath = 10,
wopAddTempRoot = 11,
wopAddIndirectRoot = 12,
wopSyncWithGC = 13,
wopFindRoots = 14,
wopExportPath = 16, // obsolete
wopQueryDeriver = 18, // obsolete
wopSetOptions = 19,
wopCollectGarbage = 20,
wopQuerySubstitutablePathInfo = 21,
wopQueryDerivationOutputs = 22, // obsolete
wopQueryAllValidPaths = 23,
wopQueryFailedPaths = 24,
wopClearFailedPaths = 25,
wopQueryPathInfo = 26,
wopImportPaths = 27, // obsolete
wopQueryDerivationOutputNames = 28, // obsolete
wopQueryPathFromHashPart = 29,
wopQuerySubstitutablePathInfos = 30,
wopQueryValidPaths = 31,
wopQuerySubstitutablePaths = 32,
wopQueryValidDerivers = 33,
wopOptimiseStore = 34,
wopVerifyStore = 35,
wopBuildDerivation = 36,
wopAddSignatures = 37,
wopNarFromPath = 38,
wopAddToStoreNar = 39,
wopQueryMissing = 40,
wopQueryDerivationOutputMap = 41,
wopRegisterDrvOutput = 42,
wopQueryRealisation = 43,
wopAddMultipleToStore = 44,
wopAddBuildLog = 45,
wopBuildPathsWithResults = 46,
} WorkerOp;
#define STDERR_NEXT 0x6f6c6d67
#define STDERR_READ 0x64617461 // data needed from source
#define STDERR_WRITE 0x64617416 // data for sink
#define STDERR_LAST 0x616c7473
#define STDERR_ERROR 0x63787470
#define STDERR_START_ACTIVITY 0x53545254
#define STDERR_STOP_ACTIVITY 0x53544f50
#define STDERR_RESULT 0x52534c54
class Store;
struct Source;
/**
* Used to guide overloading
*
* See https://en.cppreference.com/w/cpp/language/adl for the broader
* concept of what is going on here.
*/
template<typename T>
struct Phantom {};
namespace worker_proto {
/* FIXME maybe move more stuff inside here */
#define MAKE_WORKER_PROTO(TEMPLATE, T) \
TEMPLATE T read(const Store & store, Source & from, Phantom< T > _); \
TEMPLATE void write(const Store & store, Sink & out, const T & str)
MAKE_WORKER_PROTO(, std::string);
MAKE_WORKER_PROTO(, StorePath);
MAKE_WORKER_PROTO(, ContentAddress);
MAKE_WORKER_PROTO(, DerivedPath);
MAKE_WORKER_PROTO(, Realisation);
MAKE_WORKER_PROTO(, DrvOutput);
MAKE_WORKER_PROTO(, BuildResult);
MAKE_WORKER_PROTO(, KeyedBuildResult);
MAKE_WORKER_PROTO(, std::optional<TrustedFlag>);
MAKE_WORKER_PROTO(template<typename T>, std::vector<T>);
MAKE_WORKER_PROTO(template<typename T>, std::set<T>);
#define X_ template<typename K, typename V>
#define Y_ std::map<K, V>
MAKE_WORKER_PROTO(X_, Y_);
#undef X_
#undef Y_
/**
* These use the empty string for the null case, relying on the fact
* that the underlying types never serialize to the empty string.
*
* We do this instead of a generic std::optional<T> instance because
* ordinal tags (0 or 1, here) are a bit of a compatability hazard. For
* the same reason, we don't have a std::variant<T..> instances (ordinal
* tags 0...n).
*
* We could the generic instances and then these as specializations for
* compatability, but that's proven a bit finnicky, and also makes the
* worker protocol harder to implement in other languages where such
* specializations may not be allowed.
*/
MAKE_WORKER_PROTO(, std::optional<StorePath>);
MAKE_WORKER_PROTO(, std::optional<ContentAddress>);
template<typename T>
std::vector<T> read(const Store & store, Source & from, Phantom<std::vector<T>> _)
{
std::vector<T> resSet;
auto size = readNum<size_t>(from);
while (size--) {
resSet.push_back(read(store, from, Phantom<T> {}));
}
return resSet;
}
template<typename T>
void write(const Store & store, Sink & out, const std::vector<T> & resSet)
{
out << resSet.size();
for (auto & key : resSet) {
write(store, out, key);
}
}
template<typename T>
std::set<T> read(const Store & store, Source & from, Phantom<std::set<T>> _)
{
std::set<T> resSet;
auto size = readNum<size_t>(from);
while (size--) {
resSet.insert(read(store, from, Phantom<T> {}));
}
return resSet;
}
template<typename T>
void write(const Store & store, Sink & out, const std::set<T> & resSet)
{
out << resSet.size();
for (auto & key : resSet) {
write(store, out, key);
}
}
template<typename K, typename V>
std::map<K, V> read(const Store & store, Source & from, Phantom<std::map<K, V>> _)
{
std::map<K, V> resMap;
auto size = readNum<size_t>(from);
while (size--) {
auto k = read(store, from, Phantom<K> {});
auto v = read(store, from, Phantom<V> {});
resMap.insert_or_assign(std::move(k), std::move(v));
}
return resMap;
}
template<typename K, typename V>
void write(const Store & store, Sink & out, const std::map<K, V> & resMap)
{
out << resMap.size();
for (auto & i : resMap) {
write(store, out, i.first);
write(store, out, i.second);
}
}
}
}