From 7905a9f19cdb0ca2ff70567613784084e3e0a2fd Mon Sep 17 00:00:00 2001 From: Tristan Sloughter Date: Fri, 8 May 2015 20:06:43 -0500 Subject: store hex package downloads in shared cache dir --- src/rebar_pkg_resource.erl | 70 ++++++++++++++++++++++++++++++++++++++++++---- 1 file changed, 64 insertions(+), 6 deletions(-) (limited to 'src/rebar_pkg_resource.erl') diff --git a/src/rebar_pkg_resource.erl b/src/rebar_pkg_resource.erl index 3b44fc8..92e6cd4 100644 --- a/src/rebar_pkg_resource.erl +++ b/src/rebar_pkg_resource.erl @@ -9,8 +9,11 @@ ,needs_update/2 ,make_vsn/1]). +-include_lib("providers/include/providers.hrl"). -include("rebar.hrl"). +-define(DEFAULT_CDN, "https://s3.amazonaws.com/s3.hex.pm/tarballs"). + lock(_AppDir, Source) -> Source. @@ -23,12 +26,67 @@ needs_update(Dir, {pkg, _Name, Vsn}) -> true end. -download(Dir, {pkg, Name, Vsn}, State) -> - TmpFile = filename:join(Dir, "package.tar.gz"), - CDN = rebar_state:get(State, rebar_packages_cdn, "https://s3.amazonaws.com/s3.hex.pm/tarballs"), - Url = string:join([CDN, binary_to_list(<>)], "/"), - {ok, saved_to_file} = httpc:request(get, {Url, []}, [], [{stream, TmpFile}]), - {tarball, TmpFile}. +download(_Dir, {pkg, Name, Vsn}, State) -> + CDN = rebar_state:get(State, rebar_packages_cdn, ?DEFAULT_CDN), + PackageDir = hex_package_dir(CDN, State), + + Package = binary_to_list(<>), + Path = filename:join(PackageDir, Package), + Url = string:join([CDN, Package], "/"), + + case request(Url, etag(Path)) of + {ok, cached} -> + {tarball, Path}; + {ok, Binary} -> + file:write_file(Path, Binary), + {tarball, Path}; + error -> + case filelib:is_regular(Path) of + true -> + ?DEBUG("Download ~s error, using ~s since it exists", [Url, Path]), + {tarball, Path}; + false -> + throw(request_failed) + end + end. make_vsn(_) -> {error, "Replacing version of type pkg not supported."}. + +%% Use the shared hex package directory unless a non-default package repo is used +hex_package_dir(?DEFAULT_CDN, _) -> + filename:join([rebar_dir:home_dir(), ".hex", "packages"]); +hex_package_dir(CDN, State) -> + CacheDir = rebar_dir:global_cache_dir(State), + {ok, {_, _, Host, _, _, _}} = http_uri:parse(CDN), + CDNPath = filename:join(lists:reverse(string:tokens(Host, "."))), + PackageDir = filename:join([CacheDir, "hex", CDNPath, "packages"]), + ok = filelib:ensure_dir(filename:join(PackageDir, "placeholder")), + PackageDir. + +request(Url, ETag) -> + case httpc:request(get, {Url, [{"if-none-match", ETag} || ETag =/= false]}, + [{relaxed, true}], + [{body_format, binary}]) of + {ok, {{_Version, 200, _Reason}, _Headers, Body}} -> + ?DEBUG("Successfully downloaded ~s", [Url]), + {ok, Body}; + {ok, {{_Version, 304, _Reason}, _Headers, _Body}} -> + ?DEBUG("Cached copy of ~s still valid", [Url]), + {ok, cached}; + {ok, {{_Version, Code, _Reason}, _Headers, _Body}} -> + ?DEBUG("Request to ~p failed: status code ~p", [Url, Code]), + error; + {error, Reason} -> + ?DEBUG("Request to ~p failed: ~p", [Url, Reason]), + error + end. + +etag(Path) -> + case file:read_file(Path) of + {ok, Binary} -> + <> = crypto:hash(md5, Binary), + string:to_lower(lists:flatten(io_lib:format("~32.16.0b", [X]))); + {error, _} -> + false + end. -- cgit v1.1 From b636822d73410795cdc46d4722f7225c0f463fd3 Mon Sep 17 00:00:00 2001 From: Tristan Sloughter Date: Tue, 12 May 2015 09:38:22 -0500 Subject: check md5sum of package against that sent by s3 --- src/rebar_pkg_resource.erl | 17 +++++++++++++---- 1 file changed, 13 insertions(+), 4 deletions(-) (limited to 'src/rebar_pkg_resource.erl') diff --git a/src/rebar_pkg_resource.erl b/src/rebar_pkg_resource.erl index 92e6cd4..9b430d2 100644 --- a/src/rebar_pkg_resource.erl +++ b/src/rebar_pkg_resource.erl @@ -37,9 +37,17 @@ download(_Dir, {pkg, Name, Vsn}, State) -> case request(Url, etag(Path)) of {ok, cached} -> {tarball, Path}; - {ok, Binary} -> + {ok, Binary, EtagHeader} -> file:write_file(Path, Binary), - {tarball, Path}; + Etag = etag(Path), + case EtagHeader =:= Etag of + true -> + {tarball, Path}; + false -> + ?DEBUG("Bad md5sum for ~s of ~s comparing to ~s sent by server", + [Path, Etag, EtagHeader]), + throw(bad_etag) + end; error -> case filelib:is_regular(Path) of true -> @@ -68,9 +76,10 @@ request(Url, ETag) -> case httpc:request(get, {Url, [{"if-none-match", ETag} || ETag =/= false]}, [{relaxed, true}], [{body_format, binary}]) of - {ok, {{_Version, 200, _Reason}, _Headers, Body}} -> + {ok, {{_Version, 200, _Reason}, Headers, Body}} -> + {"etag", ETag1} = lists:keyfind("etag", 1, Headers), ?DEBUG("Successfully downloaded ~s", [Url]), - {ok, Body}; + {ok, Body, string:strip(ETag1, both, $")}; {ok, {{_Version, 304, _Reason}, _Headers, _Body}} -> ?DEBUG("Cached copy of ~s still valid", [Url]), {ok, cached}; -- cgit v1.1 From 56c925b75b86a6304da75635874722348ee21351 Mon Sep 17 00:00:00 2001 From: Fred Hebert Date: Tue, 12 May 2015 16:58:38 +0000 Subject: Ad-hoc attempt at restructuring pkg cache --- src/rebar_pkg_resource.erl | 79 ++++++++++++++++++++++++++++++++-------------- 1 file changed, 55 insertions(+), 24 deletions(-) (limited to 'src/rebar_pkg_resource.erl') diff --git a/src/rebar_pkg_resource.erl b/src/rebar_pkg_resource.erl index 9b430d2..56e08fc 100644 --- a/src/rebar_pkg_resource.erl +++ b/src/rebar_pkg_resource.erl @@ -26,38 +26,69 @@ needs_update(Dir, {pkg, _Name, Vsn}) -> true end. -download(_Dir, {pkg, Name, Vsn}, State) -> +download(TmpDir, Pkg={pkg, Name, Vsn}, State) -> CDN = rebar_state:get(State, rebar_packages_cdn, ?DEFAULT_CDN), PackageDir = hex_package_dir(CDN, State), - Package = binary_to_list(<>), - Path = filename:join(PackageDir, Package), + CachePath = filename:join(PackageDir, Package), Url = string:join([CDN, Package], "/"), + cached_download(TmpDir, CachePath, Pkg, Url, etag(CachePath), State). - case request(Url, etag(Path)) of +cached_download(TmpDir, CachePath, Pkg, Url, ETag, State) -> + case request(Url, ETag) of {ok, cached} -> - {tarball, Path}; - {ok, Binary, EtagHeader} -> - file:write_file(Path, Binary), - Etag = etag(Path), - case EtagHeader =:= Etag of - true -> - {tarball, Path}; - false -> - ?DEBUG("Bad md5sum for ~s of ~s comparing to ~s sent by server", - [Path, Etag, EtagHeader]), - throw(bad_etag) - end; + serve_from_cache(TmpDir, CachePath, Pkg, State); + {ok, Body, NewETag} -> + serve_from_download(TmpDir, CachePath, Pkg, NewETag, Body, State); + error when ETag =/= false -> + ?DEBUG("Download ~s error, using ~s from cache", [Url, CachePath]), + serve_from_cache(TmpDir, CachePath, Pkg, State); error -> - case filelib:is_regular(Path) of - true -> - ?DEBUG("Download ~s error, using ~s since it exists", [Url, Path]), - {tarball, Path}; - false -> - throw(request_failed) - end + throw(request_failed) + end. + +serve_from_cache(TmpDir, CachePath, Pkg, State) -> + {Files, Contents, Version, Meta} = extract(TmpDir, CachePath), + case checksums(Pkg, Files, Contents, Version, Meta, State) of + {Chk, Chk, Chk} -> + ok = erl_tar:extract({binary, Contents}, [{cwd, TmpDir}, compressed]), + {ok, true}; + {_Bin, Chk, Chk} -> + ?PRV_ERROR({failed_extract, CachePath}); + {Chk, _Reg, Chk} -> + ?PRV_ERROR({bad_registry_checksum, CachePath}); + {_Bin, _Reg, _Tar} -> + ?PRV_ERROR({bad_checksum, CachePath}) + end. + +serve_from_download(TmpDir, CachePath, Package, ETag, Binary, State) -> + ?DEBUG("Writing ~p to cache at ~s", [Package, CachePath]), + file:write_file(CachePath, Binary), + case etag(CachePath) of + ETag -> + serve_from_cache(TmpDir, CachePath, Package, State); + FileETag -> + ?DEBUG("Download ETag ~s doesn't match cached ETag ~s", [ETag, FileETag]), + ?PRV_ERROR({bad_download, CachePath}) end. + +extract(TmpDir, CachePath) -> + ec_file:mkdir_p(TmpDir), + {ok, Files} = erl_tar:extract(CachePath, [memory]), + {"contents.tar.gz", Contents} = lists:keyfind("contents.tar.gz", 1, Files), + {"VERSION", Version} = lists:keyfind("VERSION", 1, Files), + {"metadata.config", Meta} = lists:keyfind("metadata.config", 1, Files), + {Files, Contents, Version, Meta}. + +checksums(Pkg, Files, Contents, Version, Meta, State) -> + Blob = <>, + <> = crypto:hash(sha256, Blob), + BinChecksum = list_to_binary(string:to_upper(lists:flatten(io_lib:format("~64.16.0b", [X])))), + RegistryChecksum = rebar_packages:registry_sum(Pkg, State), + {"CHECKSUM", TarChecksum} = lists:keyfind("CHECKSUM", 1, Files), + {BinChecksum, RegistryChecksum, TarChecksum}. + make_vsn(_) -> {error, "Replacing version of type pkg not supported."}. @@ -77,8 +108,8 @@ request(Url, ETag) -> [{relaxed, true}], [{body_format, binary}]) of {ok, {{_Version, 200, _Reason}, Headers, Body}} -> - {"etag", ETag1} = lists:keyfind("etag", 1, Headers), ?DEBUG("Successfully downloaded ~s", [Url]), + {"etag", ETag1} = lists:keyfind("etag", 1, Headers), {ok, Body, string:strip(ETag1, both, $")}; {ok, {{_Version, 304, _Reason}, _Headers, _Body}} -> ?DEBUG("Cached copy of ~s still valid", [Url]), -- cgit v1.1 From 1643f4a1fc8a480c20ce972d9df47669111fff22 Mon Sep 17 00:00:00 2001 From: Tristan Sloughter Date: Tue, 12 May 2015 19:32:24 -0500 Subject: wrap fetch errors in rebar_fetch PRV_ERROR --- src/rebar_pkg_resource.erl | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) (limited to 'src/rebar_pkg_resource.erl') diff --git a/src/rebar_pkg_resource.erl b/src/rebar_pkg_resource.erl index 56e08fc..8bbd963 100644 --- a/src/rebar_pkg_resource.erl +++ b/src/rebar_pkg_resource.erl @@ -9,7 +9,6 @@ ,needs_update/2 ,make_vsn/1]). --include_lib("providers/include/providers.hrl"). -include("rebar.hrl"). -define(DEFAULT_CDN, "https://s3.amazonaws.com/s3.hex.pm/tarballs"). @@ -44,7 +43,7 @@ cached_download(TmpDir, CachePath, Pkg, Url, ETag, State) -> ?DEBUG("Download ~s error, using ~s from cache", [Url, CachePath]), serve_from_cache(TmpDir, CachePath, Pkg, State); error -> - throw(request_failed) + request_failed end. serve_from_cache(TmpDir, CachePath, Pkg, State) -> @@ -54,11 +53,11 @@ serve_from_cache(TmpDir, CachePath, Pkg, State) -> ok = erl_tar:extract({binary, Contents}, [{cwd, TmpDir}, compressed]), {ok, true}; {_Bin, Chk, Chk} -> - ?PRV_ERROR({failed_extract, CachePath}); + {failed_extract, CachePath}; {Chk, _Reg, Chk} -> - ?PRV_ERROR({bad_registry_checksum, CachePath}); + {bad_registry_checksum, CachePath}; {_Bin, _Reg, _Tar} -> - ?PRV_ERROR({bad_checksum, CachePath}) + {bad_checksum, CachePath} end. serve_from_download(TmpDir, CachePath, Package, ETag, Binary, State) -> @@ -69,7 +68,7 @@ serve_from_download(TmpDir, CachePath, Package, ETag, Binary, State) -> serve_from_cache(TmpDir, CachePath, Package, State); FileETag -> ?DEBUG("Download ETag ~s doesn't match cached ETag ~s", [ETag, FileETag]), - ?PRV_ERROR({bad_download, CachePath}) + {bad_download, CachePath} end. -- cgit v1.1 From 86fbaaeb08e7fc35ee87daea62be7febacd14ea9 Mon Sep 17 00:00:00 2001 From: Tristan Sloughter Date: Tue, 12 May 2015 19:36:10 -0500 Subject: add bad_download error --- src/rebar_pkg_resource.erl | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src/rebar_pkg_resource.erl') diff --git a/src/rebar_pkg_resource.erl b/src/rebar_pkg_resource.erl index 8bbd963..f6bb29b 100644 --- a/src/rebar_pkg_resource.erl +++ b/src/rebar_pkg_resource.erl @@ -67,7 +67,7 @@ serve_from_download(TmpDir, CachePath, Package, ETag, Binary, State) -> ETag -> serve_from_cache(TmpDir, CachePath, Package, State); FileETag -> - ?DEBUG("Download ETag ~s doesn't match cached ETag ~s", [ETag, FileETag]), + ?DEBUG("Download ETag ~s doesn't match returned ETag ~s", [ETag, FileETag]), {bad_download, CachePath} end. -- cgit v1.1 From ca038d39f70ea60ca4eb1c22a47e9b3bfcfe51a1 Mon Sep 17 00:00:00 2001 From: Fred Hebert Date: Wed, 13 May 2015 20:16:03 +0000 Subject: Proper custom pkg index support, some tests - The rebar package index files have been moved off the default path and will require a new `rebar3 update` - Caching of downloaded packages automatically takes place in a path relative to the CDN used - The cache path is not shared with hex as we now write and modify data in there arbitrarily - Basic tests plus the working set for more of them is included --- src/rebar_pkg_resource.erl | 26 ++++++++------------------ 1 file changed, 8 insertions(+), 18 deletions(-) (limited to 'src/rebar_pkg_resource.erl') diff --git a/src/rebar_pkg_resource.erl b/src/rebar_pkg_resource.erl index f6bb29b..59ce0dc 100644 --- a/src/rebar_pkg_resource.erl +++ b/src/rebar_pkg_resource.erl @@ -11,8 +11,6 @@ -include("rebar.hrl"). --define(DEFAULT_CDN, "https://s3.amazonaws.com/s3.hex.pm/tarballs"). - lock(_AppDir, Source) -> Source. @@ -27,7 +25,7 @@ needs_update(Dir, {pkg, _Name, Vsn}) -> download(TmpDir, Pkg={pkg, Name, Vsn}, State) -> CDN = rebar_state:get(State, rebar_packages_cdn, ?DEFAULT_CDN), - PackageDir = hex_package_dir(CDN, State), + PackageDir = rebar_packages:package_dir(State), Package = binary_to_list(<>), CachePath = filename:join(PackageDir, Package), Url = string:join([CDN, Package], "/"), @@ -53,10 +51,13 @@ serve_from_cache(TmpDir, CachePath, Pkg, State) -> ok = erl_tar:extract({binary, Contents}, [{cwd, TmpDir}, compressed]), {ok, true}; {_Bin, Chk, Chk} -> + ?DEBUG("Checksums: registry: ~p, pkg: ~p", [Chk, _Bin]), {failed_extract, CachePath}; {Chk, _Reg, Chk} -> + ?DEBUG("Checksums: registry: ~p, pkg: ~p", [_Reg, Chk]), {bad_registry_checksum, CachePath}; {_Bin, _Reg, _Tar} -> + ?DEBUG("Checksums: registry: ~p, pkg: ~p, meta: ~p", [_Reg, _Bin, _Tar]), {bad_checksum, CachePath} end. @@ -67,7 +68,7 @@ serve_from_download(TmpDir, CachePath, Package, ETag, Binary, State) -> ETag -> serve_from_cache(TmpDir, CachePath, Package, State); FileETag -> - ?DEBUG("Download ETag ~s doesn't match returned ETag ~s", [ETag, FileETag]), + ?DEBUG("Downloaded file ~s ETag ~s doesn't match returned ETag ~s", [CachePath, ETag, FileETag]), {bad_download, CachePath} end. @@ -84,28 +85,17 @@ checksums(Pkg, Files, Contents, Version, Meta, State) -> Blob = <>, <> = crypto:hash(sha256, Blob), BinChecksum = list_to_binary(string:to_upper(lists:flatten(io_lib:format("~64.16.0b", [X])))), - RegistryChecksum = rebar_packages:registry_sum(Pkg, State), + RegistryChecksum = rebar_packages:registry_checksum(Pkg, State), {"CHECKSUM", TarChecksum} = lists:keyfind("CHECKSUM", 1, Files), {BinChecksum, RegistryChecksum, TarChecksum}. make_vsn(_) -> {error, "Replacing version of type pkg not supported."}. -%% Use the shared hex package directory unless a non-default package repo is used -hex_package_dir(?DEFAULT_CDN, _) -> - filename:join([rebar_dir:home_dir(), ".hex", "packages"]); -hex_package_dir(CDN, State) -> - CacheDir = rebar_dir:global_cache_dir(State), - {ok, {_, _, Host, _, _, _}} = http_uri:parse(CDN), - CDNPath = filename:join(lists:reverse(string:tokens(Host, "."))), - PackageDir = filename:join([CacheDir, "hex", CDNPath, "packages"]), - ok = filelib:ensure_dir(filename:join(PackageDir, "placeholder")), - PackageDir. - request(Url, ETag) -> case httpc:request(get, {Url, [{"if-none-match", ETag} || ETag =/= false]}, - [{relaxed, true}], - [{body_format, binary}]) of + [{relaxed, true}], + [{body_format, binary}]) of {ok, {{_Version, 200, _Reason}, Headers, Body}} -> ?DEBUG("Successfully downloaded ~s", [Url]), {"etag", ETag1} = lists:keyfind("etag", 1, Headers), -- cgit v1.1