Issue dependencies ui kind of broken

[Aragur]

• Gitea version (or commit ref): cc9fa06
• Git version: 2.18.0
• Operating system: Debian
• Database:
  • MySQL
• Can you reproduce the bug at https://try.gitea.io:
  • Yes: https://try.gitea.io/Aragur/Issue-Dependency_Issue/issues

Description

In the dependencies list it only lists the last 8 issues relative to that one. For example issue #12 can only see issues #11 to #3. Also manual mention (for example of #1) doesn’t work.
Thanks to @Morlinest we already know more details:
Current work around is to search the exact issue name to add another.
Direct mention of the issue id is not working (but this should a thing).

Screenshots

Try to select #1 on issue #12

[jonasfranz]

Ping @kolaente

[kolaente]

This is something which should be fixed directly in the issue search, as I'm only using that (via the api). The results are the same as when you search on {projecturl}/issues.

I agree the issue search is not very good implemented.

[kolaente]

I'd rather have the search return issues by their index instead of frankensteining this into the dependencies.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs during the next 2 weeks. Thank you for your contributions.

[stevegt]

Just noticed that direct mention of the issue id is still not working as of 7d973ed.

[zeripath]

@stevegt could you give a quick example of what you mean. Preferably with a try.gitea.io link?

It might be best to open another issue since this one is closed.

[stevegt]

@zeripath A 'try' link won't help -- go to any issue you can edit and enter an issue number into the 'dependencies' field, either with or without a leading '#'. You'll see that you get an error response of 'Dependent issue does not exist.'

This issue was recently closed by stale bot due to lack of activity, not because it's been addressed. There has been no merge referencing it. Re-opening would be the normal thing to do, rather than fragment the conversation. I'll let it sit for a while to see if anyone with access re-opens. I may get to this with a fix myself at some point as well.

[nascimentolwtn]

For me is happening something similar, but it seems to be a cache problem, because if I create a sequence of issues, only old ones once shown are listed on new issues to be included as dependency. Also, it seems to show only the first page of opened issues (typing '#' would help).

A workaround: change browser or clear cache history.

[xf-]

@nascimentolwtn i think is is caused by service worker and caching the result

[lesh59]

I think the dependency drop down is only showing the first page of issues. Changing the ISSUE_PAGING_NUM in app.ini to a larger number showed more issues in the drop down (and on the issues page of course). The drop down should not be affected by "pages", should be all issues, filtered as you type.

[samaust]

Tested on version 1.11.6 :

• Searching the title of an issue that is not in the choices shown in the drop-down menu works.
• Searching the linked reference (mike/compiler#1234) does not return results.

[6543]

@samaust sorry, we dont support anymore v1.11.6 & you should upgrade for security reasons

[stevegt]

@6543 The fact that @samaust is running an old version is orthogonal to this issue -- the bug still exists in 1.14.0+dev-874-ge8ad6c1ff.

[Giszmo]

I installed from snap yesterday and ran into this same issue.

gitea 1.16.5

Not sure if bugbounties are welcome but I'd tip $50 in BTC for this to be fixed.

[Ryuno-Ki]

The relevant template is

gitea/templates/repo/issue/view_content/sidebar.tmpl

Lines 517 to 534 in 73382d2

	{{if and .CanCreateIssueDependencies (not .Repository.IsArchived)}}
	<div>
	<form method="POST" action="{{.Issue.Link}}/dependency/add" id="addDependencyForm">
	{{$.CsrfTokenHtml}}
	<div class="ui fluid action input">
	<div class="ui search selection dropdown" id="new-dependency-drop-list" data-issue-id="{{.Issue.ID}}">
	<input name="newDependency" type="hidden">
	{{svg "octicon-triangle-down" 14 "dropdown icon"}}
	<input type="text" class="search">
	<div class="default text">{{.i18n.Tr "repo.issues.dependency.add"}}</div>
	</div>
	<button class="ui green icon button">
	{{svg "octicon-plus"}}
	</button>
	</div>
	</form>
	</div>
	{{end}}

The associated JavaScript

gitea/web_src/js/features/repo-issue.js

Lines 89 to 123 in 73382d2

	export function initRepoIssueList() {
	const repolink = $('#repolink').val();
	const repoId = $('#repoId').val();
	const crossRepoSearch = $('#crossRepoSearch').val();
	const tp = $('#type').val();
	let issueSearchUrl = `${appSubUrl}/${repolink}/issues/search?q={query}&type=${tp}`;
	if (crossRepoSearch === 'true') {
	issueSearchUrl = `${appSubUrl}/issues/search?q={query}&priority_repo_id=${repoId}&type=${tp}`;
	}
	$('#new-dependency-drop-list')
	.dropdown({
	apiSettings: {
	url: issueSearchUrl,
	onResponse(response) {
	const filteredResponse = {success: true, results: []};
	const currIssueId = $('#new-dependency-drop-list').data('issue-id');
	// Parse the response from the api to work with our dropdown
	$.each(response, (_i, issue) => {
	// Don't list current issue in the dependency list.
	if (issue.id === currIssueId) {
	return;
	}
	filteredResponse.results.push({
	name: `#${issue.number} ${htmlEscape(issue.title)
	}<div class="text small dont-break-out">${htmlEscape(issue.repository.full_name)}</div>`,
	value: issue.id,
	});
	});
	return filteredResponse;
	},
	cache: false,
	},
	fullTextSearch: true,
	});

A few questions come to my mind:

1. Why is the JavaScript using .data() here? (see https://docs.gitea.io/en-us/guidelines-frontend/#html-attributes-and-dataset )
2. I can see AJAX requests like /issues/search?q=some&priority_repo_id=2&type=all&_=1654452403354 Where is the associated router? (formerly using /api/v1/repos/issues/search. Changed in 783a021#diff-d39a63ccfd5f0fe703f74ce48a3624ba59a856317b0d236d7eae87dfcf6c1c20 ) The one below?
3. Searching for something in the title or body (whole words) work. We want to reference the issue ID here, though.

gitea/routers/web/repo/search.go

Lines 19 to 58 in 7b089c4

	func Search(ctx *context.Context) {
	if !setting.Indexer.RepoIndexerEnabled {
	ctx.Redirect(ctx.Repo.RepoLink)
	return
	}
	language := ctx.FormTrim("l")
	keyword := ctx.FormTrim("q")
	page := ctx.FormInt("page")
	if page <= 0 {
	page = 1
	}
	queryType := ctx.FormTrim("t")
	isMatch := queryType == "match"
	total, searchResults, searchResultLanguages, err := code_indexer.PerformSearch(ctx, []int64{ctx.Repo.Repository.ID},
	language, keyword, page, setting.UI.RepoSearchPagingNum, isMatch)
	if err != nil {
	if code_indexer.IsAvailable() {
	ctx.ServerError("SearchResults", err)
	return
	}
	ctx.Data["CodeIndexerUnavailable"] = true
	} else {
	ctx.Data["CodeIndexerUnavailable"] = !code_indexer.IsAvailable()
	}
	ctx.Data["Keyword"] = keyword
	ctx.Data["Language"] = language
	ctx.Data["queryType"] = queryType
	ctx.Data["SourcePath"] = ctx.Repo.Repository.HTMLURL()
	ctx.Data["SearchResults"] = searchResults
	ctx.Data["SearchResultLanguages"] = searchResultLanguages
	ctx.Data["PageIsViewCode"] = true
	pager := context.NewPagination(total, setting.UI.RepoSearchPagingNum, page, 5)
	pager.SetDefaultParams(ctx)
	pager.AddParam(ctx, "l", "Language")
	ctx.Data["Page"] = pager
	ctx.HTML(http.StatusOK, tplSearch)
	}

[Ryuno-Ki]

I got an explanation on routers in the chat.

The router in

gitea/routers/web/web.go

Line 320 in d9b50e4

m.Get("/search", repo.SearchIssues)

passes the request to SearchIssues:

gitea/routers/web/repo/issue.go

Lines 2144 to 2317 in 744e452

	// SearchIssues searches for issues across the repositories that the user has access to
	func SearchIssues(ctx *context.Context) {
	before, since, err := context.GetQueryBeforeSince(ctx)
	if err != nil {
	ctx.Error(http.StatusUnprocessableEntity, err.Error())
	return
	}
	var isClosed util.OptionalBool
	switch ctx.FormString("state") {
	case "closed":
	isClosed = util.OptionalBoolTrue
	case "all":
	isClosed = util.OptionalBoolNone
	default:
	isClosed = util.OptionalBoolFalse
	}
	// find repos user can access (for issue search)
	opts := &models.SearchRepoOptions{
	Private: false,
	AllPublic: true,
	TopicOnly: false,
	Collaborate: util.OptionalBoolNone,
	// This needs to be a column that is not nil in fixtures or
	// MySQL will return different results when sorting by null in some cases
	OrderBy: db.SearchOrderByAlphabetically,
	Actor: ctx.Doer,
	}
	if ctx.IsSigned {
	opts.Private = true
	opts.AllLimited = true
	}
	if ctx.FormString("owner") != "" {
	owner, err := user_model.GetUserByName(ctx, ctx.FormString("owner"))
	if err != nil {
	if user_model.IsErrUserNotExist(err) {
	ctx.Error(http.StatusBadRequest, "Owner not found", err.Error())
	} else {
	ctx.Error(http.StatusInternalServerError, "GetUserByName", err.Error())
	}
	return
	}
	opts.OwnerID = owner.ID
	opts.AllLimited = false
	opts.AllPublic = false
	opts.Collaborate = util.OptionalBoolFalse
	}
	if ctx.FormString("team") != "" {
	if ctx.FormString("owner") == "" {
	ctx.Error(http.StatusBadRequest, "", "Owner organisation is required for filtering on team")
	return
	}
	team, err := organization.GetTeam(ctx, opts.OwnerID, ctx.FormString("team"))
	if err != nil {
	if organization.IsErrTeamNotExist(err) {
	ctx.Error(http.StatusBadRequest, "Team not found", err.Error())
	} else {
	ctx.Error(http.StatusInternalServerError, "GetUserByName", err.Error())
	}
	return
	}
	opts.TeamID = team.ID
	}
	repoCond := models.SearchRepositoryCondition(opts)
	repoIDs, _, err := models.SearchRepositoryIDs(opts)
	if err != nil {
	ctx.Error(http.StatusInternalServerError, "SearchRepositoryByName", err.Error())
	return
	}
	var issues []*models.Issue
	var filteredCount int64
	keyword := ctx.FormTrim("q")
	if strings.IndexByte(keyword, 0) >= 0 {
	keyword = ""
	}
	var issueIDs []int64
	if len(keyword) > 0 && len(repoIDs) > 0 {
	if issueIDs, err = issue_indexer.SearchIssuesByKeyword(ctx, repoIDs, keyword); err != nil {
	ctx.Error(http.StatusInternalServerError, "SearchIssuesByKeyword", err.Error())
	return
	}
	}
	var isPull util.OptionalBool
	switch ctx.FormString("type") {
	case "pulls":
	isPull = util.OptionalBoolTrue
	case "issues":
	isPull = util.OptionalBoolFalse
	default:
	isPull = util.OptionalBoolNone
	}
	labels := ctx.FormTrim("labels")
	var includedLabelNames []string
	if len(labels) > 0 {
	includedLabelNames = strings.Split(labels, ",")
	}
	milestones := ctx.FormTrim("milestones")
	var includedMilestones []string
	if len(milestones) > 0 {
	includedMilestones = strings.Split(milestones, ",")
	}
	// this api is also used in UI,
	// so the default limit is set to fit UI needs
	limit := ctx.FormInt("limit")
	if limit == 0 {
	limit = setting.UI.IssuePagingNum
	} else if limit > setting.API.MaxResponseItems {
	limit = setting.API.MaxResponseItems
	}
	// Only fetch the issues if we either don't have a keyword or the search returned issues
	// This would otherwise return all issues if no issues were found by the search.
	if len(keyword) == 0 || len(issueIDs) > 0 || len(includedLabelNames) > 0 || len(includedMilestones) > 0 {
	issuesOpt := &models.IssuesOptions{
	ListOptions: db.ListOptions{
	Page: ctx.FormInt("page"),
	PageSize: limit,
	},
	RepoCond: repoCond,
	IsClosed: isClosed,
	IssueIDs: issueIDs,
	IncludedLabelNames: includedLabelNames,
	IncludeMilestones: includedMilestones,
	SortType: "priorityrepo",
	PriorityRepoID: ctx.FormInt64("priority_repo_id"),
	IsPull: isPull,
	UpdatedBeforeUnix: before,
	UpdatedAfterUnix: since,
	}
	ctxUserID := int64(0)
	if ctx.IsSigned {
	ctxUserID = ctx.Doer.ID
	}
	// Filter for: Created by User, Assigned to User, Mentioning User, Review of User Requested
	if ctx.FormBool("created") {
	issuesOpt.PosterID = ctxUserID
	}
	if ctx.FormBool("assigned") {
	issuesOpt.AssigneeID = ctxUserID
	}
	if ctx.FormBool("mentioned") {
	issuesOpt.MentionedID = ctxUserID
	}
	if ctx.FormBool("review_requested") {
	issuesOpt.ReviewRequestedID = ctxUserID
	}
	if issues, err = models.Issues(issuesOpt); err != nil {
	ctx.Error(http.StatusInternalServerError, "Issues", err.Error())
	return
	}
	issuesOpt.ListOptions = db.ListOptions{
	Page: -1,
	}
	if filteredCount, err = models.CountIssues(issuesOpt); err != nil {
	ctx.Error(http.StatusInternalServerError, "CountIssues", err.Error())
	return
	}
	}
	ctx.SetTotalCountHeader(filteredCount)
	ctx.JSON(http.StatusOK, convert.ToAPIIssueList(issues))
	}

That makes

gitea/modules/indexer/issues/indexer.go

Lines 385 to 401 in 744e452

	func SearchIssuesByKeyword(ctx context.Context, repoIDs []int64, keyword string) ([]int64, error) {
	var issueIDs []int64
	indexer := holder.get()
	if indexer == nil {
	log.Error("SearchIssuesByKeyword(): unable to get indexer!")
	return nil, fmt.Errorf("unable to get issue indexer")
	}
	res, err := indexer.Search(ctx, keyword, repoIDs, 50, 0)
	if err != nil {
	return nil, err
	}
	for _, r := range res.Hits {
	issueIDs = append(issueIDs, r.ID)
	}
	return issueIDs, nil
	}

the code that searches the issues.

Therefore, this issue is likely not of kind/ui.

[wxiaoguang]

I can answer some of your questions.

1. Why is the JavaScript using .data() here? (see guideline )

There are a lot of legacy frontend code (as old as years ago). The rule for using attr instead of data was suggested by silverwind a few months ago and I wrote it into the guideline. For me, I am doing my best to make sure ever line of JS I touched follows the guideline.

• Add new JS linter rules #17699
• Fix some incorrect async functions, improve frontend document. #17597

2. I can see AJAX requests like .. Where is the associated router?

The easiest way is just to run the code, and watch the output log. You will see the called router function in logs (zeripath and I added this feature in 1.17, to make develop & debug more easily)

• Refactor Router Logger #17308

3. Searching for something in the title or body (whole words) work. We want to reference the issue ID here, though.

Yup, it's better to make the Issue Index can be searched or just loaded by hard-code (indeed, the number of the issue is called index internally, not id)

[d33pjs]

Having (still) the same issue. Is someone working on that? Or are there any workarounds in the meantime?

[silkentrance]

Partial matching of words is possible but might be expensive.

Bleve

Partial matching with bleve is a no brainer if you use camel case in your wording, which, of course, will not always work.

So, searching for 'iss' will never find 'issue'.
Yet, searching for 'foo' will find 'FooBar' or 'fooBar'.
The same goes for 't2', which can be found by a search for either 't' or '2'.

• Improve documentation about « partial matching » blevesearch/bleve#321 (comment)

[6543]

I propose> create a popup template that based on usage has preset filters. (e.g. only issues/pulls of repo X)

and let issues search via keyword or number have it live updated...
... multiselect would be nice
... filter for closed and open

use that:

• for adding issues/pulls in project board view to add to the project
• adding dependencys

mockup:

[silkentrance]

I propose> create a popup template that based on usage has preset filters. (e.g. only issues/pulls of repo X)

and let issues search via keyword or number have it live updated... ... multiselect would be nice ... filter for closed and open

use that:

• for adding issues/pulls in project board view to add to the project
• adding dependencys

mockup:

I propose> create a popup template that based on usage has preset filters. (e.g. only issues/pulls of repo X)

and let issues search via keyword or number have it live updated... ... multiselect would be nice ... filter for closed and open

use that:

• for adding issues/pulls in project board view to add to the project
• adding dependencys

mockup:

I strongly believe that this is an altogether different feature request.
Please make this a different issue. The issue at hand is already complex and does not require any additional complexity.

Personally, I have a different approach in mind.
One. which will move the dependency management to below the original issue/pr "description" and right before the follow-up comments, similar to for example *ira.

[silkentrance]

What is required here as a solution, is search by

• Allow searching issues by ID #31479
• any non db based indexer: fuzziness properly configured and using the proper queries on issue title
• for db only indexer: proper like query constructs on issue title