rclone/fs/cache/cache_test.go

304 lines
6.9 KiB
Go
Raw Normal View History

package cache
import (
"context"
"errors"
"testing"
"github.com/rclone/rclone/fs"
"github.com/rclone/rclone/fstest/mockfs"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
)
var (
called = 0
errSentinel = errors.New("an error")
)
func mockNewFs(t *testing.T) func(ctx context.Context, path string) (fs.Fs, error) {
called = 0
create := func(ctx context.Context, path string) (f fs.Fs, err error) {
assert.Equal(t, 0, called)
called++
switch path {
case "mock:/":
return mockfs.NewFs(ctx, "mock", "/", nil)
fs/cache: fix parent not getting pinned when remote is a file Before this change, when cache.GetFn was called on a file rather than a directory, two cache entries would be added (the file + its parent) but only one of them would get pinned if the caller then called Pin(f). This left the other one exposed to expiration if the ci.FsCacheExpireDuration was reached. This was problematic because both entries point to the same Fs, and if one entry expires while the other is pinned, the Shutdown method gets erroneously called on an Fs that is still in use. An example of the problem showed up in the Hasher backend, which uses the Shutdown method to stop the bolt db used to store hashes. If a command was run on a Hasher file (ex. `rclone md5sum --download hasher:somelargefile.zip`) and hashing the file took longer than the --fs-cache-expire-duration (5m by default), the bolt db was stopped before the hashing operation completed, resulting in an error. This change fixes the issue by ensuring that: 1. only one entry is added to the cache (the file's parent, not the file). 2. future lookups correctly find the entry regardless of whether they are called with the parent name or one of its children. 3. fs.ErrorIsFile is returned when (and only when) fsString points to a file (preserving the fix from https://github.com/rclone/rclone/commit/8d5bc7f28b8459a0e04fdbfd2dca43c5d9ec4555). Note that f.Root() should always point to the parent dir as of https://github.com/rclone/rclone/commit/c69eb84573c85206ab028eda2987180e049ef2e4
2024-03-10 05:36:57 +01:00
case "mock:/file.txt", "mock:file.txt", "mock:/file2.txt", "mock:file2.txt":
fMock, err := mockfs.NewFs(ctx, "mock", "/", nil)
require.NoError(t, err)
return fMock, fs.ErrorIsFile
case "mock:/error":
return nil, errSentinel
}
t.Fatalf("Unknown path %q", path)
panic("unreachable")
}
t.Cleanup(Clear)
return create
}
func TestGet(t *testing.T) {
create := mockNewFs(t)
assert.Equal(t, 0, Entries())
f, err := GetFn(context.Background(), "mock:/", create)
require.NoError(t, err)
assert.Equal(t, 1, Entries())
f2, err := GetFn(context.Background(), "mock:/", create)
require.NoError(t, err)
assert.Equal(t, f, f2)
}
func TestGetFile(t *testing.T) {
fs/cache: fix parent not getting pinned when remote is a file Before this change, when cache.GetFn was called on a file rather than a directory, two cache entries would be added (the file + its parent) but only one of them would get pinned if the caller then called Pin(f). This left the other one exposed to expiration if the ci.FsCacheExpireDuration was reached. This was problematic because both entries point to the same Fs, and if one entry expires while the other is pinned, the Shutdown method gets erroneously called on an Fs that is still in use. An example of the problem showed up in the Hasher backend, which uses the Shutdown method to stop the bolt db used to store hashes. If a command was run on a Hasher file (ex. `rclone md5sum --download hasher:somelargefile.zip`) and hashing the file took longer than the --fs-cache-expire-duration (5m by default), the bolt db was stopped before the hashing operation completed, resulting in an error. This change fixes the issue by ensuring that: 1. only one entry is added to the cache (the file's parent, not the file). 2. future lookups correctly find the entry regardless of whether they are called with the parent name or one of its children. 3. fs.ErrorIsFile is returned when (and only when) fsString points to a file (preserving the fix from https://github.com/rclone/rclone/commit/8d5bc7f28b8459a0e04fdbfd2dca43c5d9ec4555). Note that f.Root() should always point to the parent dir as of https://github.com/rclone/rclone/commit/c69eb84573c85206ab028eda2987180e049ef2e4
2024-03-10 05:36:57 +01:00
defer ClearMappings()
create := mockNewFs(t)
assert.Equal(t, 0, Entries())
f, err := GetFn(context.Background(), "mock:/file.txt", create)
require.Equal(t, fs.ErrorIsFile, err)
require.NotNil(t, f)
fs/cache: fix parent not getting pinned when remote is a file Before this change, when cache.GetFn was called on a file rather than a directory, two cache entries would be added (the file + its parent) but only one of them would get pinned if the caller then called Pin(f). This left the other one exposed to expiration if the ci.FsCacheExpireDuration was reached. This was problematic because both entries point to the same Fs, and if one entry expires while the other is pinned, the Shutdown method gets erroneously called on an Fs that is still in use. An example of the problem showed up in the Hasher backend, which uses the Shutdown method to stop the bolt db used to store hashes. If a command was run on a Hasher file (ex. `rclone md5sum --download hasher:somelargefile.zip`) and hashing the file took longer than the --fs-cache-expire-duration (5m by default), the bolt db was stopped before the hashing operation completed, resulting in an error. This change fixes the issue by ensuring that: 1. only one entry is added to the cache (the file's parent, not the file). 2. future lookups correctly find the entry regardless of whether they are called with the parent name or one of its children. 3. fs.ErrorIsFile is returned when (and only when) fsString points to a file (preserving the fix from https://github.com/rclone/rclone/commit/8d5bc7f28b8459a0e04fdbfd2dca43c5d9ec4555). Note that f.Root() should always point to the parent dir as of https://github.com/rclone/rclone/commit/c69eb84573c85206ab028eda2987180e049ef2e4
2024-03-10 05:36:57 +01:00
assert.Equal(t, 1, Entries())
f2, err := GetFn(context.Background(), "mock:/file.txt", create)
require.Equal(t, fs.ErrorIsFile, err)
require.NotNil(t, f2)
assert.Equal(t, f, f2)
fs/cache: fix parent not getting pinned when remote is a file Before this change, when cache.GetFn was called on a file rather than a directory, two cache entries would be added (the file + its parent) but only one of them would get pinned if the caller then called Pin(f). This left the other one exposed to expiration if the ci.FsCacheExpireDuration was reached. This was problematic because both entries point to the same Fs, and if one entry expires while the other is pinned, the Shutdown method gets erroneously called on an Fs that is still in use. An example of the problem showed up in the Hasher backend, which uses the Shutdown method to stop the bolt db used to store hashes. If a command was run on a Hasher file (ex. `rclone md5sum --download hasher:somelargefile.zip`) and hashing the file took longer than the --fs-cache-expire-duration (5m by default), the bolt db was stopped before the hashing operation completed, resulting in an error. This change fixes the issue by ensuring that: 1. only one entry is added to the cache (the file's parent, not the file). 2. future lookups correctly find the entry regardless of whether they are called with the parent name or one of its children. 3. fs.ErrorIsFile is returned when (and only when) fsString points to a file (preserving the fix from https://github.com/rclone/rclone/commit/8d5bc7f28b8459a0e04fdbfd2dca43c5d9ec4555). Note that f.Root() should always point to the parent dir as of https://github.com/rclone/rclone/commit/c69eb84573c85206ab028eda2987180e049ef2e4
2024-03-10 05:36:57 +01:00
// check it is also found when referred to by parent name
f2, err = GetFn(context.Background(), "mock:/", create)
require.Nil(t, err)
require.NotNil(t, f2)
assert.Equal(t, f, f2)
}
func TestGetFile2(t *testing.T) {
fs/cache: fix parent not getting pinned when remote is a file Before this change, when cache.GetFn was called on a file rather than a directory, two cache entries would be added (the file + its parent) but only one of them would get pinned if the caller then called Pin(f). This left the other one exposed to expiration if the ci.FsCacheExpireDuration was reached. This was problematic because both entries point to the same Fs, and if one entry expires while the other is pinned, the Shutdown method gets erroneously called on an Fs that is still in use. An example of the problem showed up in the Hasher backend, which uses the Shutdown method to stop the bolt db used to store hashes. If a command was run on a Hasher file (ex. `rclone md5sum --download hasher:somelargefile.zip`) and hashing the file took longer than the --fs-cache-expire-duration (5m by default), the bolt db was stopped before the hashing operation completed, resulting in an error. This change fixes the issue by ensuring that: 1. only one entry is added to the cache (the file's parent, not the file). 2. future lookups correctly find the entry regardless of whether they are called with the parent name or one of its children. 3. fs.ErrorIsFile is returned when (and only when) fsString points to a file (preserving the fix from https://github.com/rclone/rclone/commit/8d5bc7f28b8459a0e04fdbfd2dca43c5d9ec4555). Note that f.Root() should always point to the parent dir as of https://github.com/rclone/rclone/commit/c69eb84573c85206ab028eda2987180e049ef2e4
2024-03-10 05:36:57 +01:00
defer ClearMappings()
create := mockNewFs(t)
assert.Equal(t, 0, Entries())
f, err := GetFn(context.Background(), "mock:file.txt", create)
require.Equal(t, fs.ErrorIsFile, err)
require.NotNil(t, f)
fs/cache: fix parent not getting pinned when remote is a file Before this change, when cache.GetFn was called on a file rather than a directory, two cache entries would be added (the file + its parent) but only one of them would get pinned if the caller then called Pin(f). This left the other one exposed to expiration if the ci.FsCacheExpireDuration was reached. This was problematic because both entries point to the same Fs, and if one entry expires while the other is pinned, the Shutdown method gets erroneously called on an Fs that is still in use. An example of the problem showed up in the Hasher backend, which uses the Shutdown method to stop the bolt db used to store hashes. If a command was run on a Hasher file (ex. `rclone md5sum --download hasher:somelargefile.zip`) and hashing the file took longer than the --fs-cache-expire-duration (5m by default), the bolt db was stopped before the hashing operation completed, resulting in an error. This change fixes the issue by ensuring that: 1. only one entry is added to the cache (the file's parent, not the file). 2. future lookups correctly find the entry regardless of whether they are called with the parent name or one of its children. 3. fs.ErrorIsFile is returned when (and only when) fsString points to a file (preserving the fix from https://github.com/rclone/rclone/commit/8d5bc7f28b8459a0e04fdbfd2dca43c5d9ec4555). Note that f.Root() should always point to the parent dir as of https://github.com/rclone/rclone/commit/c69eb84573c85206ab028eda2987180e049ef2e4
2024-03-10 05:36:57 +01:00
assert.Equal(t, 1, Entries())
f2, err := GetFn(context.Background(), "mock:file.txt", create)
require.Equal(t, fs.ErrorIsFile, err)
require.NotNil(t, f2)
assert.Equal(t, f, f2)
fs/cache: fix parent not getting pinned when remote is a file Before this change, when cache.GetFn was called on a file rather than a directory, two cache entries would be added (the file + its parent) but only one of them would get pinned if the caller then called Pin(f). This left the other one exposed to expiration if the ci.FsCacheExpireDuration was reached. This was problematic because both entries point to the same Fs, and if one entry expires while the other is pinned, the Shutdown method gets erroneously called on an Fs that is still in use. An example of the problem showed up in the Hasher backend, which uses the Shutdown method to stop the bolt db used to store hashes. If a command was run on a Hasher file (ex. `rclone md5sum --download hasher:somelargefile.zip`) and hashing the file took longer than the --fs-cache-expire-duration (5m by default), the bolt db was stopped before the hashing operation completed, resulting in an error. This change fixes the issue by ensuring that: 1. only one entry is added to the cache (the file's parent, not the file). 2. future lookups correctly find the entry regardless of whether they are called with the parent name or one of its children. 3. fs.ErrorIsFile is returned when (and only when) fsString points to a file (preserving the fix from https://github.com/rclone/rclone/commit/8d5bc7f28b8459a0e04fdbfd2dca43c5d9ec4555). Note that f.Root() should always point to the parent dir as of https://github.com/rclone/rclone/commit/c69eb84573c85206ab028eda2987180e049ef2e4
2024-03-10 05:36:57 +01:00
// check it is also found when referred to by parent name
f2, err = GetFn(context.Background(), "mock:/", create)
require.Nil(t, err)
require.NotNil(t, f2)
assert.Equal(t, f, f2)
}
func TestGetError(t *testing.T) {
create := mockNewFs(t)
assert.Equal(t, 0, Entries())
f, err := GetFn(context.Background(), "mock:/error", create)
require.Equal(t, errSentinel, err)
require.Equal(t, nil, f)
assert.Equal(t, 0, Entries())
}
func TestPutErr(t *testing.T) {
create := mockNewFs(t)
f, err := mockfs.NewFs(context.Background(), "mock", "", nil)
require.NoError(t, err)
assert.Equal(t, 0, Entries())
fs/cache: fix parent not getting pinned when remote is a file Before this change, when cache.GetFn was called on a file rather than a directory, two cache entries would be added (the file + its parent) but only one of them would get pinned if the caller then called Pin(f). This left the other one exposed to expiration if the ci.FsCacheExpireDuration was reached. This was problematic because both entries point to the same Fs, and if one entry expires while the other is pinned, the Shutdown method gets erroneously called on an Fs that is still in use. An example of the problem showed up in the Hasher backend, which uses the Shutdown method to stop the bolt db used to store hashes. If a command was run on a Hasher file (ex. `rclone md5sum --download hasher:somelargefile.zip`) and hashing the file took longer than the --fs-cache-expire-duration (5m by default), the bolt db was stopped before the hashing operation completed, resulting in an error. This change fixes the issue by ensuring that: 1. only one entry is added to the cache (the file's parent, not the file). 2. future lookups correctly find the entry regardless of whether they are called with the parent name or one of its children. 3. fs.ErrorIsFile is returned when (and only when) fsString points to a file (preserving the fix from https://github.com/rclone/rclone/commit/8d5bc7f28b8459a0e04fdbfd2dca43c5d9ec4555). Note that f.Root() should always point to the parent dir as of https://github.com/rclone/rclone/commit/c69eb84573c85206ab028eda2987180e049ef2e4
2024-03-10 05:36:57 +01:00
PutErr("mock:/", f, fs.ErrorNotFoundInConfigFile)
assert.Equal(t, 1, Entries())
fs/cache: fix parent not getting pinned when remote is a file Before this change, when cache.GetFn was called on a file rather than a directory, two cache entries would be added (the file + its parent) but only one of them would get pinned if the caller then called Pin(f). This left the other one exposed to expiration if the ci.FsCacheExpireDuration was reached. This was problematic because both entries point to the same Fs, and if one entry expires while the other is pinned, the Shutdown method gets erroneously called on an Fs that is still in use. An example of the problem showed up in the Hasher backend, which uses the Shutdown method to stop the bolt db used to store hashes. If a command was run on a Hasher file (ex. `rclone md5sum --download hasher:somelargefile.zip`) and hashing the file took longer than the --fs-cache-expire-duration (5m by default), the bolt db was stopped before the hashing operation completed, resulting in an error. This change fixes the issue by ensuring that: 1. only one entry is added to the cache (the file's parent, not the file). 2. future lookups correctly find the entry regardless of whether they are called with the parent name or one of its children. 3. fs.ErrorIsFile is returned when (and only when) fsString points to a file (preserving the fix from https://github.com/rclone/rclone/commit/8d5bc7f28b8459a0e04fdbfd2dca43c5d9ec4555). Note that f.Root() should always point to the parent dir as of https://github.com/rclone/rclone/commit/c69eb84573c85206ab028eda2987180e049ef2e4
2024-03-10 05:36:57 +01:00
fNew, err := GetFn(context.Background(), "mock:/", create)
require.Equal(t, fs.ErrorNotFoundInConfigFile, err)
require.Equal(t, f, fNew)
assert.Equal(t, 1, Entries())
// Check canonicalisation
fs/cache: fix parent not getting pinned when remote is a file Before this change, when cache.GetFn was called on a file rather than a directory, two cache entries would be added (the file + its parent) but only one of them would get pinned if the caller then called Pin(f). This left the other one exposed to expiration if the ci.FsCacheExpireDuration was reached. This was problematic because both entries point to the same Fs, and if one entry expires while the other is pinned, the Shutdown method gets erroneously called on an Fs that is still in use. An example of the problem showed up in the Hasher backend, which uses the Shutdown method to stop the bolt db used to store hashes. If a command was run on a Hasher file (ex. `rclone md5sum --download hasher:somelargefile.zip`) and hashing the file took longer than the --fs-cache-expire-duration (5m by default), the bolt db was stopped before the hashing operation completed, resulting in an error. This change fixes the issue by ensuring that: 1. only one entry is added to the cache (the file's parent, not the file). 2. future lookups correctly find the entry regardless of whether they are called with the parent name or one of its children. 3. fs.ErrorIsFile is returned when (and only when) fsString points to a file (preserving the fix from https://github.com/rclone/rclone/commit/8d5bc7f28b8459a0e04fdbfd2dca43c5d9ec4555). Note that f.Root() should always point to the parent dir as of https://github.com/rclone/rclone/commit/c69eb84573c85206ab028eda2987180e049ef2e4
2024-03-10 05:36:57 +01:00
PutErr("mock:/file.txt", f, fs.ErrorNotFoundInConfigFile)
fNew, err = GetFn(context.Background(), "mock:/file.txt", create)
fs/cache: fix parent not getting pinned when remote is a file Before this change, when cache.GetFn was called on a file rather than a directory, two cache entries would be added (the file + its parent) but only one of them would get pinned if the caller then called Pin(f). This left the other one exposed to expiration if the ci.FsCacheExpireDuration was reached. This was problematic because both entries point to the same Fs, and if one entry expires while the other is pinned, the Shutdown method gets erroneously called on an Fs that is still in use. An example of the problem showed up in the Hasher backend, which uses the Shutdown method to stop the bolt db used to store hashes. If a command was run on a Hasher file (ex. `rclone md5sum --download hasher:somelargefile.zip`) and hashing the file took longer than the --fs-cache-expire-duration (5m by default), the bolt db was stopped before the hashing operation completed, resulting in an error. This change fixes the issue by ensuring that: 1. only one entry is added to the cache (the file's parent, not the file). 2. future lookups correctly find the entry regardless of whether they are called with the parent name or one of its children. 3. fs.ErrorIsFile is returned when (and only when) fsString points to a file (preserving the fix from https://github.com/rclone/rclone/commit/8d5bc7f28b8459a0e04fdbfd2dca43c5d9ec4555). Note that f.Root() should always point to the parent dir as of https://github.com/rclone/rclone/commit/c69eb84573c85206ab028eda2987180e049ef2e4
2024-03-10 05:36:57 +01:00
require.Equal(t, fs.ErrorNotFoundInConfigFile, err)
require.Equal(t, f, fNew)
assert.Equal(t, 1, Entries())
}
func TestPut(t *testing.T) {
create := mockNewFs(t)
f, err := mockfs.NewFs(context.Background(), "mock", "/alien", nil)
require.NoError(t, err)
assert.Equal(t, 0, Entries())
Put("mock:/alien", f)
assert.Equal(t, 1, Entries())
fNew, err := GetFn(context.Background(), "mock:/alien", create)
require.NoError(t, err)
require.Equal(t, f, fNew)
assert.Equal(t, 1, Entries())
// Check canonicalisation
Put("mock:/alien/", f)
fNew, err = GetFn(context.Background(), "mock:/alien/", create)
require.NoError(t, err)
require.Equal(t, f, fNew)
assert.Equal(t, 1, Entries())
}
func TestPin(t *testing.T) {
create := mockNewFs(t)
// Test pinning and unpinning nonexistent
f, err := mockfs.NewFs(context.Background(), "mock", "/alien", nil)
require.NoError(t, err)
Pin(f)
Unpin(f)
// Now test pinning an existing
f2, err := GetFn(context.Background(), "mock:/", create)
require.NoError(t, err)
Pin(f2)
Unpin(f2)
}
fs/cache: fix parent not getting pinned when remote is a file Before this change, when cache.GetFn was called on a file rather than a directory, two cache entries would be added (the file + its parent) but only one of them would get pinned if the caller then called Pin(f). This left the other one exposed to expiration if the ci.FsCacheExpireDuration was reached. This was problematic because both entries point to the same Fs, and if one entry expires while the other is pinned, the Shutdown method gets erroneously called on an Fs that is still in use. An example of the problem showed up in the Hasher backend, which uses the Shutdown method to stop the bolt db used to store hashes. If a command was run on a Hasher file (ex. `rclone md5sum --download hasher:somelargefile.zip`) and hashing the file took longer than the --fs-cache-expire-duration (5m by default), the bolt db was stopped before the hashing operation completed, resulting in an error. This change fixes the issue by ensuring that: 1. only one entry is added to the cache (the file's parent, not the file). 2. future lookups correctly find the entry regardless of whether they are called with the parent name or one of its children. 3. fs.ErrorIsFile is returned when (and only when) fsString points to a file (preserving the fix from https://github.com/rclone/rclone/commit/8d5bc7f28b8459a0e04fdbfd2dca43c5d9ec4555). Note that f.Root() should always point to the parent dir as of https://github.com/rclone/rclone/commit/c69eb84573c85206ab028eda2987180e049ef2e4
2024-03-10 05:36:57 +01:00
func TestPinFile(t *testing.T) {
defer ClearMappings()
create := mockNewFs(t)
// Test pinning and unpinning nonexistent
f, err := mockfs.NewFs(context.Background(), "mock", "/file.txt", nil)
require.NoError(t, err)
Pin(f)
Unpin(f)
// Now test pinning an existing
f2, err := GetFn(context.Background(), "mock:/file.txt", create)
require.Equal(t, fs.ErrorIsFile, err)
assert.Equal(t, 1, len(childParentMap))
Pin(f2)
assert.Equal(t, 1, Entries())
pinned, unpinned := EntriesWithPinCount()
assert.Equal(t, 1, pinned)
assert.Equal(t, 0, unpinned)
Unpin(f2)
assert.Equal(t, 1, Entries())
pinned, unpinned = EntriesWithPinCount()
assert.Equal(t, 0, pinned)
assert.Equal(t, 1, unpinned)
// try a different child of the same parent, and parent
// should not add additional cache items
called = 0 // this one does create() because we haven't seen it before and don't yet know it's a file
f3, err := GetFn(context.Background(), "mock:/file2.txt", create)
assert.Equal(t, fs.ErrorIsFile, err)
assert.Equal(t, 1, Entries())
assert.Equal(t, 2, len(childParentMap))
parent, err := GetFn(context.Background(), "mock:/", create)
assert.NoError(t, err)
assert.Equal(t, 1, Entries())
assert.Equal(t, 2, len(childParentMap))
Pin(f3)
assert.Equal(t, 1, Entries())
pinned, unpinned = EntriesWithPinCount()
assert.Equal(t, 1, pinned)
assert.Equal(t, 0, unpinned)
Unpin(f3)
assert.Equal(t, 1, Entries())
pinned, unpinned = EntriesWithPinCount()
assert.Equal(t, 0, pinned)
assert.Equal(t, 1, unpinned)
Pin(parent)
assert.Equal(t, 1, Entries())
pinned, unpinned = EntriesWithPinCount()
assert.Equal(t, 1, pinned)
assert.Equal(t, 0, unpinned)
Unpin(parent)
assert.Equal(t, 1, Entries())
pinned, unpinned = EntriesWithPinCount()
assert.Equal(t, 0, pinned)
assert.Equal(t, 1, unpinned)
// all 3 should have equal configstrings
assert.Equal(t, fs.ConfigString(f2), fs.ConfigString(f3))
assert.Equal(t, fs.ConfigString(f2), fs.ConfigString(parent))
}
func TestClearConfig(t *testing.T) {
create := mockNewFs(t)
assert.Equal(t, 0, Entries())
_, err := GetFn(context.Background(), "mock:/file.txt", create)
require.Equal(t, fs.ErrorIsFile, err)
fs/cache: fix parent not getting pinned when remote is a file Before this change, when cache.GetFn was called on a file rather than a directory, two cache entries would be added (the file + its parent) but only one of them would get pinned if the caller then called Pin(f). This left the other one exposed to expiration if the ci.FsCacheExpireDuration was reached. This was problematic because both entries point to the same Fs, and if one entry expires while the other is pinned, the Shutdown method gets erroneously called on an Fs that is still in use. An example of the problem showed up in the Hasher backend, which uses the Shutdown method to stop the bolt db used to store hashes. If a command was run on a Hasher file (ex. `rclone md5sum --download hasher:somelargefile.zip`) and hashing the file took longer than the --fs-cache-expire-duration (5m by default), the bolt db was stopped before the hashing operation completed, resulting in an error. This change fixes the issue by ensuring that: 1. only one entry is added to the cache (the file's parent, not the file). 2. future lookups correctly find the entry regardless of whether they are called with the parent name or one of its children. 3. fs.ErrorIsFile is returned when (and only when) fsString points to a file (preserving the fix from https://github.com/rclone/rclone/commit/8d5bc7f28b8459a0e04fdbfd2dca43c5d9ec4555). Note that f.Root() should always point to the parent dir as of https://github.com/rclone/rclone/commit/c69eb84573c85206ab028eda2987180e049ef2e4
2024-03-10 05:36:57 +01:00
assert.Equal(t, 1, Entries())
fs/cache: fix parent not getting pinned when remote is a file Before this change, when cache.GetFn was called on a file rather than a directory, two cache entries would be added (the file + its parent) but only one of them would get pinned if the caller then called Pin(f). This left the other one exposed to expiration if the ci.FsCacheExpireDuration was reached. This was problematic because both entries point to the same Fs, and if one entry expires while the other is pinned, the Shutdown method gets erroneously called on an Fs that is still in use. An example of the problem showed up in the Hasher backend, which uses the Shutdown method to stop the bolt db used to store hashes. If a command was run on a Hasher file (ex. `rclone md5sum --download hasher:somelargefile.zip`) and hashing the file took longer than the --fs-cache-expire-duration (5m by default), the bolt db was stopped before the hashing operation completed, resulting in an error. This change fixes the issue by ensuring that: 1. only one entry is added to the cache (the file's parent, not the file). 2. future lookups correctly find the entry regardless of whether they are called with the parent name or one of its children. 3. fs.ErrorIsFile is returned when (and only when) fsString points to a file (preserving the fix from https://github.com/rclone/rclone/commit/8d5bc7f28b8459a0e04fdbfd2dca43c5d9ec4555). Note that f.Root() should always point to the parent dir as of https://github.com/rclone/rclone/commit/c69eb84573c85206ab028eda2987180e049ef2e4
2024-03-10 05:36:57 +01:00
assert.Equal(t, 1, ClearConfig("mock"))
assert.Equal(t, 0, Entries())
}
func TestClear(t *testing.T) {
create := mockNewFs(t)
// Create something
_, err := GetFn(context.Background(), "mock:/", create)
require.NoError(t, err)
assert.Equal(t, 1, Entries())
Clear()
assert.Equal(t, 0, Entries())
}
func TestEntries(t *testing.T) {
create := mockNewFs(t)
assert.Equal(t, 0, Entries())
// Create something
_, err := GetFn(context.Background(), "mock:/", create)
require.NoError(t, err)
assert.Equal(t, 1, Entries())
}