guix/nix/libstore/misc.cc

221 lines
7 KiB
C++
Raw Normal View History

#include "misc.hh"
#include "store-api.hh"
#include "local-store.hh"
#include "globals.hh"
namespace nix {
Derivation derivationFromPath(StoreAPI & store, const Path & drvPath)
{
assertStorePath(drvPath);
store.ensurePath(drvPath);
return readDerivation(drvPath);
}
void computeFSClosure(StoreAPI & store, const Path & path,
PathSet & paths, bool flipDirection, bool includeOutputs, bool includeDerivers)
{
if (paths.find(path) != paths.end()) return;
paths.insert(path);
PathSet edges;
if (flipDirection) {
store.queryReferrers(path, edges);
if (includeOutputs) {
PathSet derivers = store.queryValidDerivers(path);
foreach (PathSet::iterator, i, derivers)
edges.insert(*i);
}
if (includeDerivers && isDerivation(path)) {
PathSet outputs = store.queryDerivationOutputs(path);
foreach (PathSet::iterator, i, outputs)
if (store.isValidPath(*i) && store.queryDeriver(*i) == path)
edges.insert(*i);
}
} else {
store.queryReferences(path, edges);
if (includeOutputs && isDerivation(path)) {
PathSet outputs = store.queryDerivationOutputs(path);
foreach (PathSet::iterator, i, outputs)
if (store.isValidPath(*i)) edges.insert(*i);
}
if (includeDerivers) {
Path deriver = store.queryDeriver(path);
if (store.isValidPath(deriver)) edges.insert(deriver);
}
}
foreach (PathSet::iterator, i, edges)
computeFSClosure(store, *i, paths, flipDirection, includeOutputs, includeDerivers);
}
Path findOutput(const Derivation & drv, string id)
{
2009-04-21 11:52:16 +00:00
foreach (DerivationOutputs::const_iterator, i, drv.outputs)
if (i->first == id) return i->second.path;
throw Error(format("derivation has no output `%1%'") % id);
}
void queryMissing(StoreAPI & store, const PathSet & targets,
PathSet & willBuild, PathSet & willSubstitute, PathSet & unknown,
unsigned long long & downloadSize, unsigned long long & narSize)
{
downloadSize = narSize = 0;
PathSet todo(targets.begin(), targets.end()), done;
download-from-binary-cache: parallelise fetching of NAR info files Getting substitute information using the binary cache substituter has non-trivial latency overhead. A package or NixOS system configuration can have hundreds of dependencies, and in the worst case (when the local info cache is empty) we have to do a separate HTTP request for each of these. If the ping time to the server is t, getting N info files will take tN seconds; e.g., with a ping time of 0.1s to nixos.org, sequentially downloading 1000 info files (a typical NixOS config) will take at least 100 seconds. To fix this problem, the binary cache substituter can now perform requests in parallel. This required changing the substituter interface to support a function querySubstitutablePathInfos() that queries multiple paths at the same time, and rewriting queryMissing() to take advantage of parallelism. (Due to local caching, parallelising queryMissing() is sufficient for most use cases, since it's almost always called before building a derivation and thus fills the local info cache.) For example, parallelism speeds up querying all 1056 paths in a particular NixOS system configuration from 116s to 2.6s. It works so well because the eccentricity of the top-level derivation in the dependency graph is only 9. So we only need 10 round-trips (when using an unlimited number of parallel connections) to get everything. Currently we do a maximum of 150 parallel connections to the server. Thus it's important that the binary cache server (e.g. nixos.org) has a high connection limit. Alternatively we could use HTTP pipelining, but WWW::Curl doesn't support it and libcurl has a hard-coded limit of 5 requests per pipeline.
2012-07-06 23:08:20 +00:00
/* Getting substitute info has high latency when using the binary
cache substituter. Thus it's essential to do substitute
queries in parallel as much as possible. To accomplish this
we do the following:
- For all paths still to be processed (todo), we add all
paths for which we need info to the set query. For an
unbuilt derivation this is the output paths; otherwise, it's
the path itself.
- We get info about all paths in query in parallel.
- We process the results and add new items to todo if
necessary. E.g. if a path is substitutable, then we need to
get info on its references.
- Repeat until todo is empty.
*/
while (!todo.empty()) {
download-from-binary-cache: parallelise fetching of NAR info files Getting substitute information using the binary cache substituter has non-trivial latency overhead. A package or NixOS system configuration can have hundreds of dependencies, and in the worst case (when the local info cache is empty) we have to do a separate HTTP request for each of these. If the ping time to the server is t, getting N info files will take tN seconds; e.g., with a ping time of 0.1s to nixos.org, sequentially downloading 1000 info files (a typical NixOS config) will take at least 100 seconds. To fix this problem, the binary cache substituter can now perform requests in parallel. This required changing the substituter interface to support a function querySubstitutablePathInfos() that queries multiple paths at the same time, and rewriting queryMissing() to take advantage of parallelism. (Due to local caching, parallelising queryMissing() is sufficient for most use cases, since it's almost always called before building a derivation and thus fills the local info cache.) For example, parallelism speeds up querying all 1056 paths in a particular NixOS system configuration from 116s to 2.6s. It works so well because the eccentricity of the top-level derivation in the dependency graph is only 9. So we only need 10 round-trips (when using an unlimited number of parallel connections) to get everything. Currently we do a maximum of 150 parallel connections to the server. Thus it's important that the binary cache server (e.g. nixos.org) has a high connection limit. Alternatively we could use HTTP pipelining, but WWW::Curl doesn't support it and libcurl has a hard-coded limit of 5 requests per pipeline.
2012-07-06 23:08:20 +00:00
PathSet query, todoDrv, todoNonDrv;
foreach (PathSet::iterator, i, todo) {
if (done.find(*i) != done.end()) continue;
done.insert(*i);
DrvPathWithOutputs i2 = parseDrvPathWithOutputs(*i);
if (isDerivation(i2.first)) {
if (!store.isValidPath(i2.first)) {
download-from-binary-cache: parallelise fetching of NAR info files Getting substitute information using the binary cache substituter has non-trivial latency overhead. A package or NixOS system configuration can have hundreds of dependencies, and in the worst case (when the local info cache is empty) we have to do a separate HTTP request for each of these. If the ping time to the server is t, getting N info files will take tN seconds; e.g., with a ping time of 0.1s to nixos.org, sequentially downloading 1000 info files (a typical NixOS config) will take at least 100 seconds. To fix this problem, the binary cache substituter can now perform requests in parallel. This required changing the substituter interface to support a function querySubstitutablePathInfos() that queries multiple paths at the same time, and rewriting queryMissing() to take advantage of parallelism. (Due to local caching, parallelising queryMissing() is sufficient for most use cases, since it's almost always called before building a derivation and thus fills the local info cache.) For example, parallelism speeds up querying all 1056 paths in a particular NixOS system configuration from 116s to 2.6s. It works so well because the eccentricity of the top-level derivation in the dependency graph is only 9. So we only need 10 round-trips (when using an unlimited number of parallel connections) to get everything. Currently we do a maximum of 150 parallel connections to the server. Thus it's important that the binary cache server (e.g. nixos.org) has a high connection limit. Alternatively we could use HTTP pipelining, but WWW::Curl doesn't support it and libcurl has a hard-coded limit of 5 requests per pipeline.
2012-07-06 23:08:20 +00:00
// FIXME: we could try to substitute p.
unknown.insert(*i);
continue;
}
Derivation drv = derivationFromPath(store, i2.first);
download-from-binary-cache: parallelise fetching of NAR info files Getting substitute information using the binary cache substituter has non-trivial latency overhead. A package or NixOS system configuration can have hundreds of dependencies, and in the worst case (when the local info cache is empty) we have to do a separate HTTP request for each of these. If the ping time to the server is t, getting N info files will take tN seconds; e.g., with a ping time of 0.1s to nixos.org, sequentially downloading 1000 info files (a typical NixOS config) will take at least 100 seconds. To fix this problem, the binary cache substituter can now perform requests in parallel. This required changing the substituter interface to support a function querySubstitutablePathInfos() that queries multiple paths at the same time, and rewriting queryMissing() to take advantage of parallelism. (Due to local caching, parallelising queryMissing() is sufficient for most use cases, since it's almost always called before building a derivation and thus fills the local info cache.) For example, parallelism speeds up querying all 1056 paths in a particular NixOS system configuration from 116s to 2.6s. It works so well because the eccentricity of the top-level derivation in the dependency graph is only 9. So we only need 10 round-trips (when using an unlimited number of parallel connections) to get everything. Currently we do a maximum of 150 parallel connections to the server. Thus it's important that the binary cache server (e.g. nixos.org) has a high connection limit. Alternatively we could use HTTP pipelining, but WWW::Curl doesn't support it and libcurl has a hard-coded limit of 5 requests per pipeline.
2012-07-06 23:08:20 +00:00
PathSet invalid;
foreach (DerivationOutputs::iterator, j, drv.outputs)
if (wantOutput(j->first, i2.second)
&& !store.isValidPath(j->second.path))
invalid.insert(j->second.path);
download-from-binary-cache: parallelise fetching of NAR info files Getting substitute information using the binary cache substituter has non-trivial latency overhead. A package or NixOS system configuration can have hundreds of dependencies, and in the worst case (when the local info cache is empty) we have to do a separate HTTP request for each of these. If the ping time to the server is t, getting N info files will take tN seconds; e.g., with a ping time of 0.1s to nixos.org, sequentially downloading 1000 info files (a typical NixOS config) will take at least 100 seconds. To fix this problem, the binary cache substituter can now perform requests in parallel. This required changing the substituter interface to support a function querySubstitutablePathInfos() that queries multiple paths at the same time, and rewriting queryMissing() to take advantage of parallelism. (Due to local caching, parallelising queryMissing() is sufficient for most use cases, since it's almost always called before building a derivation and thus fills the local info cache.) For example, parallelism speeds up querying all 1056 paths in a particular NixOS system configuration from 116s to 2.6s. It works so well because the eccentricity of the top-level derivation in the dependency graph is only 9. So we only need 10 round-trips (when using an unlimited number of parallel connections) to get everything. Currently we do a maximum of 150 parallel connections to the server. Thus it's important that the binary cache server (e.g. nixos.org) has a high connection limit. Alternatively we could use HTTP pipelining, but WWW::Curl doesn't support it and libcurl has a hard-coded limit of 5 requests per pipeline.
2012-07-06 23:08:20 +00:00
if (invalid.empty()) continue;
todoDrv.insert(*i);
if (settings.useSubstitutes && !willBuildLocally(drv))
query.insert(invalid.begin(), invalid.end());
download-from-binary-cache: parallelise fetching of NAR info files Getting substitute information using the binary cache substituter has non-trivial latency overhead. A package or NixOS system configuration can have hundreds of dependencies, and in the worst case (when the local info cache is empty) we have to do a separate HTTP request for each of these. If the ping time to the server is t, getting N info files will take tN seconds; e.g., with a ping time of 0.1s to nixos.org, sequentially downloading 1000 info files (a typical NixOS config) will take at least 100 seconds. To fix this problem, the binary cache substituter can now perform requests in parallel. This required changing the substituter interface to support a function querySubstitutablePathInfos() that queries multiple paths at the same time, and rewriting queryMissing() to take advantage of parallelism. (Due to local caching, parallelising queryMissing() is sufficient for most use cases, since it's almost always called before building a derivation and thus fills the local info cache.) For example, parallelism speeds up querying all 1056 paths in a particular NixOS system configuration from 116s to 2.6s. It works so well because the eccentricity of the top-level derivation in the dependency graph is only 9. So we only need 10 round-trips (when using an unlimited number of parallel connections) to get everything. Currently we do a maximum of 150 parallel connections to the server. Thus it's important that the binary cache server (e.g. nixos.org) has a high connection limit. Alternatively we could use HTTP pipelining, but WWW::Curl doesn't support it and libcurl has a hard-coded limit of 5 requests per pipeline.
2012-07-06 23:08:20 +00:00
}
else {
if (store.isValidPath(*i)) continue;
query.insert(*i);
todoNonDrv.insert(*i);
}
download-from-binary-cache: parallelise fetching of NAR info files Getting substitute information using the binary cache substituter has non-trivial latency overhead. A package or NixOS system configuration can have hundreds of dependencies, and in the worst case (when the local info cache is empty) we have to do a separate HTTP request for each of these. If the ping time to the server is t, getting N info files will take tN seconds; e.g., with a ping time of 0.1s to nixos.org, sequentially downloading 1000 info files (a typical NixOS config) will take at least 100 seconds. To fix this problem, the binary cache substituter can now perform requests in parallel. This required changing the substituter interface to support a function querySubstitutablePathInfos() that queries multiple paths at the same time, and rewriting queryMissing() to take advantage of parallelism. (Due to local caching, parallelising queryMissing() is sufficient for most use cases, since it's almost always called before building a derivation and thus fills the local info cache.) For example, parallelism speeds up querying all 1056 paths in a particular NixOS system configuration from 116s to 2.6s. It works so well because the eccentricity of the top-level derivation in the dependency graph is only 9. So we only need 10 round-trips (when using an unlimited number of parallel connections) to get everything. Currently we do a maximum of 150 parallel connections to the server. Thus it's important that the binary cache server (e.g. nixos.org) has a high connection limit. Alternatively we could use HTTP pipelining, but WWW::Curl doesn't support it and libcurl has a hard-coded limit of 5 requests per pipeline.
2012-07-06 23:08:20 +00:00
}
todo.clear();
download-from-binary-cache: parallelise fetching of NAR info files Getting substitute information using the binary cache substituter has non-trivial latency overhead. A package or NixOS system configuration can have hundreds of dependencies, and in the worst case (when the local info cache is empty) we have to do a separate HTTP request for each of these. If the ping time to the server is t, getting N info files will take tN seconds; e.g., with a ping time of 0.1s to nixos.org, sequentially downloading 1000 info files (a typical NixOS config) will take at least 100 seconds. To fix this problem, the binary cache substituter can now perform requests in parallel. This required changing the substituter interface to support a function querySubstitutablePathInfos() that queries multiple paths at the same time, and rewriting queryMissing() to take advantage of parallelism. (Due to local caching, parallelising queryMissing() is sufficient for most use cases, since it's almost always called before building a derivation and thus fills the local info cache.) For example, parallelism speeds up querying all 1056 paths in a particular NixOS system configuration from 116s to 2.6s. It works so well because the eccentricity of the top-level derivation in the dependency graph is only 9. So we only need 10 round-trips (when using an unlimited number of parallel connections) to get everything. Currently we do a maximum of 150 parallel connections to the server. Thus it's important that the binary cache server (e.g. nixos.org) has a high connection limit. Alternatively we could use HTTP pipelining, but WWW::Curl doesn't support it and libcurl has a hard-coded limit of 5 requests per pipeline.
2012-07-06 23:08:20 +00:00
SubstitutablePathInfos infos;
store.querySubstitutablePathInfos(query, infos);
foreach (PathSet::iterator, i, todoDrv) {
DrvPathWithOutputs i2 = parseDrvPathWithOutputs(*i);
download-from-binary-cache: parallelise fetching of NAR info files Getting substitute information using the binary cache substituter has non-trivial latency overhead. A package or NixOS system configuration can have hundreds of dependencies, and in the worst case (when the local info cache is empty) we have to do a separate HTTP request for each of these. If the ping time to the server is t, getting N info files will take tN seconds; e.g., with a ping time of 0.1s to nixos.org, sequentially downloading 1000 info files (a typical NixOS config) will take at least 100 seconds. To fix this problem, the binary cache substituter can now perform requests in parallel. This required changing the substituter interface to support a function querySubstitutablePathInfos() that queries multiple paths at the same time, and rewriting queryMissing() to take advantage of parallelism. (Due to local caching, parallelising queryMissing() is sufficient for most use cases, since it's almost always called before building a derivation and thus fills the local info cache.) For example, parallelism speeds up querying all 1056 paths in a particular NixOS system configuration from 116s to 2.6s. It works so well because the eccentricity of the top-level derivation in the dependency graph is only 9. So we only need 10 round-trips (when using an unlimited number of parallel connections) to get everything. Currently we do a maximum of 150 parallel connections to the server. Thus it's important that the binary cache server (e.g. nixos.org) has a high connection limit. Alternatively we could use HTTP pipelining, but WWW::Curl doesn't support it and libcurl has a hard-coded limit of 5 requests per pipeline.
2012-07-06 23:08:20 +00:00
// FIXME: cache this
Derivation drv = derivationFromPath(store, i2.first);
PathSet outputs;
bool mustBuild = false;
if (settings.useSubstitutes && !willBuildLocally(drv)) {
foreach (DerivationOutputs::iterator, j, drv.outputs) {
if (!wantOutput(j->first, i2.second)) continue;
if (!store.isValidPath(j->second.path)) {
if (infos.find(j->second.path) == infos.end())
mustBuild = true;
else
outputs.insert(j->second.path);
}
}
download-from-binary-cache: parallelise fetching of NAR info files Getting substitute information using the binary cache substituter has non-trivial latency overhead. A package or NixOS system configuration can have hundreds of dependencies, and in the worst case (when the local info cache is empty) we have to do a separate HTTP request for each of these. If the ping time to the server is t, getting N info files will take tN seconds; e.g., with a ping time of 0.1s to nixos.org, sequentially downloading 1000 info files (a typical NixOS config) will take at least 100 seconds. To fix this problem, the binary cache substituter can now perform requests in parallel. This required changing the substituter interface to support a function querySubstitutablePathInfos() that queries multiple paths at the same time, and rewriting queryMissing() to take advantage of parallelism. (Due to local caching, parallelising queryMissing() is sufficient for most use cases, since it's almost always called before building a derivation and thus fills the local info cache.) For example, parallelism speeds up querying all 1056 paths in a particular NixOS system configuration from 116s to 2.6s. It works so well because the eccentricity of the top-level derivation in the dependency graph is only 9. So we only need 10 round-trips (when using an unlimited number of parallel connections) to get everything. Currently we do a maximum of 150 parallel connections to the server. Thus it's important that the binary cache server (e.g. nixos.org) has a high connection limit. Alternatively we could use HTTP pipelining, but WWW::Curl doesn't support it and libcurl has a hard-coded limit of 5 requests per pipeline.
2012-07-06 23:08:20 +00:00
} else
mustBuild = true;
if (mustBuild) {
willBuild.insert(i2.first);
todo.insert(drv.inputSrcs.begin(), drv.inputSrcs.end());
foreach (DerivationInputs::iterator, j, drv.inputDrvs)
todo.insert(makeDrvPathWithOutputs(j->first, j->second));
download-from-binary-cache: parallelise fetching of NAR info files Getting substitute information using the binary cache substituter has non-trivial latency overhead. A package or NixOS system configuration can have hundreds of dependencies, and in the worst case (when the local info cache is empty) we have to do a separate HTTP request for each of these. If the ping time to the server is t, getting N info files will take tN seconds; e.g., with a ping time of 0.1s to nixos.org, sequentially downloading 1000 info files (a typical NixOS config) will take at least 100 seconds. To fix this problem, the binary cache substituter can now perform requests in parallel. This required changing the substituter interface to support a function querySubstitutablePathInfos() that queries multiple paths at the same time, and rewriting queryMissing() to take advantage of parallelism. (Due to local caching, parallelising queryMissing() is sufficient for most use cases, since it's almost always called before building a derivation and thus fills the local info cache.) For example, parallelism speeds up querying all 1056 paths in a particular NixOS system configuration from 116s to 2.6s. It works so well because the eccentricity of the top-level derivation in the dependency graph is only 9. So we only need 10 round-trips (when using an unlimited number of parallel connections) to get everything. Currently we do a maximum of 150 parallel connections to the server. Thus it's important that the binary cache server (e.g. nixos.org) has a high connection limit. Alternatively we could use HTTP pipelining, but WWW::Curl doesn't support it and libcurl has a hard-coded limit of 5 requests per pipeline.
2012-07-06 23:08:20 +00:00
} else
todoNonDrv.insert(outputs.begin(), outputs.end());
}
download-from-binary-cache: parallelise fetching of NAR info files Getting substitute information using the binary cache substituter has non-trivial latency overhead. A package or NixOS system configuration can have hundreds of dependencies, and in the worst case (when the local info cache is empty) we have to do a separate HTTP request for each of these. If the ping time to the server is t, getting N info files will take tN seconds; e.g., with a ping time of 0.1s to nixos.org, sequentially downloading 1000 info files (a typical NixOS config) will take at least 100 seconds. To fix this problem, the binary cache substituter can now perform requests in parallel. This required changing the substituter interface to support a function querySubstitutablePathInfos() that queries multiple paths at the same time, and rewriting queryMissing() to take advantage of parallelism. (Due to local caching, parallelising queryMissing() is sufficient for most use cases, since it's almost always called before building a derivation and thus fills the local info cache.) For example, parallelism speeds up querying all 1056 paths in a particular NixOS system configuration from 116s to 2.6s. It works so well because the eccentricity of the top-level derivation in the dependency graph is only 9. So we only need 10 round-trips (when using an unlimited number of parallel connections) to get everything. Currently we do a maximum of 150 parallel connections to the server. Thus it's important that the binary cache server (e.g. nixos.org) has a high connection limit. Alternatively we could use HTTP pipelining, but WWW::Curl doesn't support it and libcurl has a hard-coded limit of 5 requests per pipeline.
2012-07-06 23:08:20 +00:00
foreach (PathSet::iterator, i, todoNonDrv) {
done.insert(*i);
SubstitutablePathInfos::iterator info = infos.find(*i);
if (info != infos.end()) {
willSubstitute.insert(*i);
downloadSize += info->second.downloadSize;
narSize += info->second.narSize;
todo.insert(info->second.references.begin(), info->second.references.end());
} else
download-from-binary-cache: parallelise fetching of NAR info files Getting substitute information using the binary cache substituter has non-trivial latency overhead. A package or NixOS system configuration can have hundreds of dependencies, and in the worst case (when the local info cache is empty) we have to do a separate HTTP request for each of these. If the ping time to the server is t, getting N info files will take tN seconds; e.g., with a ping time of 0.1s to nixos.org, sequentially downloading 1000 info files (a typical NixOS config) will take at least 100 seconds. To fix this problem, the binary cache substituter can now perform requests in parallel. This required changing the substituter interface to support a function querySubstitutablePathInfos() that queries multiple paths at the same time, and rewriting queryMissing() to take advantage of parallelism. (Due to local caching, parallelising queryMissing() is sufficient for most use cases, since it's almost always called before building a derivation and thus fills the local info cache.) For example, parallelism speeds up querying all 1056 paths in a particular NixOS system configuration from 116s to 2.6s. It works so well because the eccentricity of the top-level derivation in the dependency graph is only 9. So we only need 10 round-trips (when using an unlimited number of parallel connections) to get everything. Currently we do a maximum of 150 parallel connections to the server. Thus it's important that the binary cache server (e.g. nixos.org) has a high connection limit. Alternatively we could use HTTP pipelining, but WWW::Curl doesn't support it and libcurl has a hard-coded limit of 5 requests per pipeline.
2012-07-06 23:08:20 +00:00
unknown.insert(*i);
}
}
}
2011-12-30 15:02:50 +00:00
static void dfsVisit(StoreAPI & store, const PathSet & paths,
const Path & path, PathSet & visited, Paths & sorted,
PathSet & parents)
{
if (parents.find(path) != parents.end())
throw BuildError(format("cycle detected in the references of `%1%'") % path);
2011-12-30 15:02:50 +00:00
if (visited.find(path) != visited.end()) return;
visited.insert(path);
2011-12-30 17:13:25 +00:00
parents.insert(path);
2011-12-30 15:02:50 +00:00
PathSet references;
if (store.isValidPath(path))
store.queryReferences(path, references);
2011-12-30 15:02:50 +00:00
foreach (PathSet::iterator, i, references)
/* Don't traverse into paths that don't exist. That can
happen due to substitutes for non-existent paths. */
if (*i != path && paths.find(*i) != paths.end())
dfsVisit(store, paths, *i, visited, sorted, parents);
sorted.push_front(path);
parents.erase(path);
}
Paths topoSortPaths(StoreAPI & store, const PathSet & paths)
{
Paths sorted;
PathSet visited, parents;
foreach (PathSet::const_iterator, i, paths)
dfsVisit(store, paths, *i, visited, sorted, parents);
return sorted;
}
}